<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Joy Rathnayake on Medium]]></title>
        <description><![CDATA[Stories by Joy Rathnayake on Medium]]></description>
        <link>https://medium.com/@jrathnayake?source=rss-50cfe99c97f3------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*srz2Qu3e5baN9Z51uXKcVg.jpeg</url>
            <title>Stories by Joy Rathnayake on Medium</title>
            <link>https://medium.com/@jrathnayake?source=rss-50cfe99c97f3------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 23:49:21 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@jrathnayake/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[API Gateway vs AI Gateway: They Solve Different Problems]]></title>
            <link>https://jrathnayake.medium.com/api-gateway-vs-ai-gateway-they-solve-different-problems-53e27b29e4e6?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/53e27b29e4e6</guid>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[api-gateway]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai-gateway]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Mon, 26 Oct 2026 08:32:16 GMT</pubDate>
            <atom:updated>2026-10-26T08:32:16.861Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Evolution of Gateway Architecture</h3><p>Modern microservice architectures rely on <strong><em>API Gateways</em></strong> as standard ingress control points for security, rate limiting, and fast routing of predictable, low-latency requests. Placing an API Gateway in front of internal services is an established industry standard for enforcing perimeter security and managing traffic.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3TNukQqy46bC2kcNZehxmQ.png" /><figcaption>Traditional API Gateway</figcaption></figure><p>Integrating large language models (LLMs) fundamentally alters these traffic patterns. Rather than exchanging lightweight, structured payloads in milliseconds, LLM calls handle large prompt contexts and long-running, streaming responses that can span seconds or minutes. Traditional API Gateways were not architected to govern long-running, conversational traffic, nor can they manage the cost structures, security risks, and operational scale associated with LLM workloads. Managing these requirements effectively necessitates a dedicated egress control layer: the <strong><em>AI Gateway</em></strong>.</p><h3>Architectural Incompatibilities of Traditional Gateways</h3><p>Attempting to govern LLM traffic by simply routing requests through an existing API Gateway introduces fundamental operational failures. Traditional API Gateway policies are designed for deterministic web traffic patterns that do not apply to AI workloads:</p><ul><li><strong>Latency and Connection Timeouts:</strong> Standard web APIs process small data payloads in milliseconds. AI models process large text contexts and take seconds or minutes to compute a response.</li></ul><blockquote><em>Example: a user asks an AI to summarize a 50-page document. The model takes 45 seconds to read and generate the summary. A traditional Gateway might cut the connection at 30 seconds, assuming the server crashed.</em></blockquote><ul><li><strong>Streaming vs. Request-Response Mechanics:</strong> Traditional APIs rely on synchronous request-and-response patterns. AI workloads typically stream data incrementally via Server-Sent Events (SSE) or WebSockets. Buffering gateways introduce significant latency by attempting to collect the entire response body before forwarding it to the client.</li></ul><blockquote><em>Example: when someone chats with an AI assistant, they expect to see the answer typed out in real time. A standard API Gateway often tries to buffer the data and send the whole response at once, which creates a frustrating delay.</em></blockquote><ul><li><strong>Statefulness and Conversational Context:</strong> Standard API routing is stateless, treating every incoming request as an isolated event. Generative AI interactions require maintaining conversational history across multiple requests to ensure contextual continuity.</li></ul><blockquote><em>Example: if someone asks, “What is the capital of France?” and follows up with, “What’s the weather like there today?”, the AI needs the context of the first question to answer the second. Traditional Gateways don’t track that ongoing conversational state.</em></blockquote><ul><li><strong>Request Volume vs. Consumption-Based Metrics: </strong>Traditional rate-limiting caps request counts per time window. In LLM workloads, request volume correlates poorly with resource consumption or financial cost.</li></ul><blockquote><em>Example: take two calls to the same model. The first classifies a support ticket: maybe 40 tokens in, 10 out. The second summarizes a 90-page contract: 60,000 tokens in, 2,000 out. To a standard Gateway, those are two requests. To the other teams, they differ by three orders of magnitude. So when a platform team caps a business unit at 10,000 calls a day, the next question should be what that cap is worth in money/ expenses, and nobody has ever been able to answer it. The cap isn’t wrong exactly. It’s just measuring the wrong thing, which is worse, because it looks like in control.</em></blockquote><h3>The Inversion of Traffic Direction: Ingress vs. Egress</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*O4-XDJDi684k3tBIFf-90g.png" /><figcaption>Inversion of Traffic Direction: Ingress vs. Egress</figcaption></figure><p>Beyond execution mechanics, traditional API Gateways and AI Gateways differ in traffic direction and trust boundaries.</p><p><em>An API Gateway governs ingress</em>: untrusted external clients accessing internal microservices. Its primary security posture answers: “Is this client authorized to enter the network?”</p><p><em>An AI Gateway governs egress</em>: internal applications reaching out to third-party model providers or external infrastructure. Its primary security posture answers: “What data is leaving our boundary, and where is it being sent?”</p><p>This direction reversal creates three primary architectural risks:</p><ol><li><strong>Vendor Key Sprawl and Blast Radius</strong></li></ol><p>Model provider API keys act as bearer tokens tied directly to enterprise billing accounts. Storing these static credentials directly in application repositories, CI/CD variables, or local development environments creates severe security exposure. Revoking a compromised global key causes widespread service outages across all consuming applications. An AI Gateway centralizes these credentials, exposing scoped, internally managed identity tokens to application services instead.</p><p><strong>2. Unmonitored Data Exfiltration</strong></p><p>Standard Data Loss Prevention (DLP) tools monitor file transfers and endpoint egress, but often fail to inspect HTTPS POST payloads containing contextual user data sent to external LLM endpoints. Without active outbound payload inspection, sensitive customer data or intellectual property can breach data residency and regulatory compliance frameworks (e.g., GDPR, HIPAA).</p><p><strong>3. Cascading Availability and Quota Failures</strong></p><p>When an internal service fails, teams can scale, roll back, or patch the deployment. When an external model provider experiences degraded performance or enforces account-level rate limits (HTTP 429), all applications sharing that provider account are throttled simultaneously. A background batch process can easily deplete an account quota, starving real-time, customer-facing applications without showing errors on internal service metrics.</p><p><strong><em>An AI Gateway is not an API Gateway with plugins; ingress and egress control points fulfill distinct operational mandates.</em></strong></p><h3>Core Operational Responsibilities of an AI Gateway</h3><p>Operating as a centralized control layer for outbound traffic, an AI Gateway delivers several fundamental capabilities, four of which prove particularly essential:</p><ol><li><strong>Inline Cost and Spend Control</strong></li></ol><p>Effective cost governance begins with token-aware rate limiting. Rather than enforcing static request counts, an AI Gateway measures traffic throughput based on input prompt and output completion tokens. Standard HTTP request caps fail to reflect actual resource usage, whereas token-aware metering accurately maps to compute loads and financial expenditure generated by large context windows.</p><p>To prevent unexpected budget overruns, active blocking and circuit breakers intercept requests before they cross the perimeter once defined cost thresholds are reached. Soft limits issue automated alerts at specific budget percentages to allow graceful model switching or degraded service levels, while hard circuit breakers instantly reject outgoing calls to prevent runaway billing loops.</p><p>Finally, cost attribution and chargeback analytics ensure financial accountability across teams. By automatically injecting and capturing structured metadata, such as application IDs, team identities, and cost centers, the AI Gateway provides precise telemetry for spend tracking. This enables automated chargebacks, FinOps reporting, and clear audit trails without requiring changes to client application code.</p><p><strong>2. Routing, so applications stop naming vendors</strong></p><p>Applications should ask for a capability, not an endpoint: something like document summarizer, resolved by the gateway to whichever provider and model currently sits behind that label.</p><p>The immediate payoff is failover, and that’s usually how it gets sold. The payoff that matters more shows up at the six-month mark, when a provider retires a model. They do this on their own schedule, and the notice period is generous but finite. If 50 applications have a model identifier compiled into them, that’s 50 tickets, 50 regression cycles, and one very unhappy quarter. Alternatively, if they all point at a logical name, it’s one configuration change and a canary experience.</p><p>Tiered Inference / cost-tiered routing operates in this same layer. Not every prompt requires a top-tier model; routing simple queries to lower-cost models saves significant budget. However, this requires careful testing to avoid subtle quality regressions.</p><p><strong>3. Guardrails, applied in both directions</strong></p><p>Inbound, before the prompt leaves our perimeter — strip or tokenize personal data, and screen for injection attempts. Outbound, before the response reaches the user — check for leaked system instructions, internal data, and content our policy doesn’t allow.</p><p>Guardrails are probabilistic, not absolute. Classifiers can produce false positives, and strict injection checks may block valid user requests while adding latency. The reason to place guardrails at the Gateway is to ensure consistent, centrally managed inspection across all outbound AI traffic.</p><p><strong>4. Governance, which is mostly a data problem</strong></p><p>Governance relies on two core elements: <strong><em>policy</em></strong> (defining which teams access specific models, regions, and data classifications) and <strong><em>evidence</em></strong> (maintaining an audit trail of requests, responses, and gateway actions, with sensitive fields redacted at write time).</p><p>An explicit audit trail is essential for compliance frameworks like the EU AI Act or ISO 42001. Answering audit questions should require a simple query against gateway logs rather than tracing code across dozens of repositories.</p><p>Observability metrics also differ from standard web APIs. Rather than focusing solely on requests per second or basic error rates, key metrics include tokens per team, cost per model, time-to-first-token, and cache hit rates.</p><p>These requirements align with standard gateway goals: controlling consumption, abstracting backends, inspecting traffic, and auditing behaviour. Only the underlying units of measurement have changed.</p><h3>Where the AI Gateway sits in our architecture</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KBeyDTuAEmrDqB00ilZbZw.png" /><figcaption>API Gateway and AI Gateway together</figcaption></figure><blockquote><strong>Does an AI Gateway replace an API Gateway?</strong></blockquote><blockquote><strong>No. They handle distinct operational responsibilities.</strong></blockquote><p>Inbound business APIs keep going through the API Gateway exactly as they do today: partner integrations, mobile backends, internal service traffic. Nothing about that changes because you started using LLMs.</p><p>Outbound model traffic goes through the AI Gateway. Applications call it instead of calling LLM providers directly, and it holds the provider credentials, applies quota, routes to a model, and writes the audit trail.</p><blockquote>Both sit on the same identity layer. This is the piece people leave off their diagrams and then rebuild badly six months later. The token our application presents to the AI Gateway should come from the same issuer that secures everything else we run, not a second, parallel authentication scheme that exists only for model traffic, which is what we get by default when the AI work grows inside one project team.</blockquote><p>When an application receives a request through an API Gateway, processes it, and calls a LLM via an AI Gateway, trace context must persist across both control points to ensure log correlation during debugging.</p><h3>When Agents start calling Tools.?</h3><p>Everything above assumes an application calling a LLM. That assumption is already loosening.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_RxEFE0NWuIcEMwymtF4qg.png" /><figcaption>When Agent start calling Tools</figcaption></figure><p>Agents call tools. Tools are APIs, which means agent traffic arrives at our existing API Gateway, without a human anywhere in the request. The Model Context Protocol (MCP) has given this some structure, and MCP servers are increasingly the shape that tool access takes. That raises questions the current generation of gateways doesn’t answer well.</p><blockquote><strong><em>What identity does an agent hold? How do we express that it’s acting on behalf of a specific user, with that user’s permissions and not more? What happens when it calls the same tool 400 times in a loop?</em></strong></blockquote><p>Nobody has finished answering those questions yet. But the shape of the answer looks familiar: a policy enforcement point between the caller and the destination it wants to reach, doing identity, quota, and audit.</p><p>To resolve these emerging agentic challenges, an AI Gateway extends its control capabilities in three key ways:</p><ul><li><strong><em>Agent Identity &amp; Auth Context:</em></strong><em> </em>Issues dedicated, short-lived service identities for agents while attaching verifiable metadata to establish trace context across multi-step execution flows.</li><li><strong><em>User Permission Delegation:</em></strong> Enforces scoped delegation protocols (such as OAuth 2.0 Token Exchange or On-Behalf-Of flows) so that tool invocation bounds strictly respect the initiating user’s privileges.</li><li><strong><em>Loop Detection &amp; Call Limits: </em></strong>Implements dynamic recursion detection and per-session depth limits to break automated execution loops and prevent run-away financial or infrastructure resource consumption.</li></ul><h3>When Agents start calling other Agents (A2A)?</h3><p>Tool calls are the first step. The next one is already happening: agents delegating work to other agents. The Agent2Agent (A2A) protocol, which Google introduced in 2025 and has since handed to the Linux Foundation, gives this a standard shape. MCP connects an Agent to Tools and Data. A2A connects an Agent to another Agent, often one built by a different team, on a different framework, or owned by a different company.</p><p>Each agent publishes an Agent Card, a JSON document that describes what it can do, where its endpoint is, and which authentication schemes it accepts. A client agent reads the card, sends a task, and gets results back, often as a long-running task that streams updates over Server-Sent Events. Those are exactly the traffic patterns that broke traditional API Gateways we discussed earlier — long-lived connections, streamed responses, and work that spans many requests.</p><blockquote>Example: a customer asks a support agent about a disputed charge. The support agent delegates to a billing agent, which asks a fraud-scoring agent for a risk check, which calls a partner’s verification agent. That’s four agents, three trust boundaries and possibly two companies, and the human who started it is now three hops back.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wElFwM6G6NTmkZ5HvUEC3w.png" /><figcaption>When Agents start calling other Agents (A2A)</figcaption></figure><p>That chain raises questions a single API call never did:</p><ul><li><strong>Whose authority is this?</strong> By the third hop, does the fraud agent know it’s acting for that specific customer, with that customer’s permissions, or only that “the billing agent asked”?</li><li><strong>Who is this agent, really? </strong>An Agent Card is a self-description. Anyone can publish one that claims to do “payments reconciliation.”</li><li><strong>What did it pass along? </strong>Each agent’s output becomes the next agent’s input. A prompt injection planted in one agent’s response can steer every agent downstream.</li><li><strong>What did it cost?</strong> Each agent in the chain may call its own models and tools. A single user request can fan out into dozens of LLM calls that nobody budgeted for.</li><li><strong>What happened?</strong> When the answer is wrong, can we rebuild the whole chain, or only the first hop?</li></ul><h4>Why A2A traffic belongs behind an AI Gateway</h4><p>The answer has the same shape as before: a policy enforcement point between the caller and what it wants to reach. For A2A, that means the AI Gateway takes on six jobs:</p><ol><li><strong>A governed agent registry.</strong> Agents should discover each other through a curated catalog, not by fetching any Agent Card they find. The AI Gateway decides which agents are approved, which versions, and which skills each consumer may invoke. It’s the same thing an API Developer Portal does for APIs.</li><li><strong>Agent identity, not just agent claims.</strong> Every agent gets a verifiable workload identity from the same issuer that secures the rest of our estate. Calls between agents are mutually authenticated, so a card that says “billing agent” has to be backed by credentials that prove it.</li><li><strong>Delegation that survives the chain.</strong> The gateway exchanges tokens at each hop (for example, OAuth 2.0 Token Exchange), narrowing scope as it goes. The fraud agent receives a token that says “acting for this customer, read-only, for this task,” not a copy of the billing agent’s broad service credentials.</li><li><strong>Inspection in both directions.</strong> Agent-to-agent payloads get the same guardrails as model traffic: PII redaction before data crosses a boundary, and treating every agent’s response as untrusted input before it reaches the next agent.</li><li><strong>Chain-level limits.</strong> Per-call rate limits don’t catch a runaway delegation tree. The gateway tracks depth, fan-out and token spend across the whole task, and stops it when any of them crosses a threshold.</li><li><strong>One audit trail for the whole conversation.</strong> Trace context carries through every hop, so “why did the system decline this refund?” is one query across gateway logs, not an investigation across four teams’ services.</li></ol><p><strong><em>Without an AI Gateway, A2A turns into a mesh of agents trusting each other’s self-descriptions, passing broad credentials down the chain, and spending money nobody can trace. With an AI Gateway, every hop is identified, scoped, inspected, metered and logged. That’s what lets an organization move from a few experimental agents to agents that collaborate across teams and companies without losing control of who did what, for whom, and at what cost.</em></strong></p><h3>Architectural Anti-Patterns</h3><p>When implementing AI Gateway patterns, avoid these common organizational pitfalls:</p><ul><li><strong>Distributing Provider SDKs directly across applications:</strong> Linking vendor SDKs directly into microservices creates tight coupling. Upgrading libraries or applying global security policies then requires synchronized, multi-repository code deployments across independent software delivery teams.</li><li><strong>Treating basic reverse proxies as AI Gateways:</strong> Standard reverse proxies manage basic token pass-through but lack native capabilities for token parsing, streaming response inspection, and semantic caching. Extending standard proxies with custom scripts quickly results in a fragile, unmaintainable in-house proxy platform.</li><li><strong>Decentralized, team-level implementations:</strong> Allowing teams to deploy custom gateway solutions leads to fragmented credential storage, inconsistent audit logs, and an inability to enforce enterprise-wide spend governance.</li></ul><h3>A quick-reference comparison</h3><p>AI Gateways complement existing API Gateways by managing a distinct outbound control point. The most effective approach is consolidating model traffic through a centralized AI Gateway while adopting capability-based routing.</p><p>Here’s a quick summary of how the core responsibilities change when handling AI workloads:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HwV_VrZWtAnmg6MDdhm7GQ.png" /><figcaption>API Gateway vs AI Gateway</figcaption></figure><p>This doesn’t make an API Gateway a bad product. It is just built for a different kind of traffic. <strong>Token-aware quotas</strong>, <strong>semantic caching</strong>, and <strong>streaming-aware mediation</strong> aren’t settings we tune, they require different underlying machinery.</p><h3>Conclusion</h3><p>API management practices evolved over years to standardize enterprise ingress control. Enterprise adoption of LLMs has occurred at a much faster pace, often leaving security, cost, and compliance controls to be backfitted post-deployment.</p><p>Establishing a centralized egress control layer via an AI Gateway enables organizations to secure model interactions, bound operational expenditure, and maintain auditability without compromising application development speed.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=53e27b29e4e6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[DPDP Act in Plain English — Decoding India’s Data Protection Act for the Rest of Us]]></title>
            <link>https://jrathnayake.medium.com/dpdp-act-in-plain-english-decoding-indias-data-protection-act-78b9c99d7ab6?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/78b9c99d7ab6</guid>
            <category><![CDATA[india]]></category>
            <category><![CDATA[dpdp-act]]></category>
            <category><![CDATA[compliance]]></category>
            <category><![CDATA[wso2]]></category>
            <category><![CDATA[regulatory-compliance]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Sat, 04 Jul 2026 11:14:46 GMT</pubDate>
            <atom:updated>2026-08-20T05:36:48.997Z</atom:updated>
            <content:encoded><![CDATA[<h3>DPDP Act in Plain English — Decoding India’s Data Protection Act for the Rest of Us</h3><p><em>The Everyday Citizen’s Guide to the DPDP Act — </em>What Happens When You Flip the Toggle?</p><p>Tired of wading through endless technical jargon trying to understand the DPDP Act and its impact on your life? Here is The Everyday Citizen’s Guide to the DPDP Act, your clear, jargon-free roadmap.</p><h3>1. Introduction- Legislative Overview &amp; Jurisdictional Scope</h3><p>The <strong>Digital Personal Data Protection (DPDP) Act, 2023</strong> represents India’s first standalone legislative framework dedicated entirely to informational privacy. Prior to its enactment, India lacked a singular, centralized data protection authority or a comprehensive privacy code. Instead, digital privacy was handled via a patchwork of generic cyber law provisions and sectoral regulations.</p><p>With the notification of the <strong>DPDP Rules, 2025</strong>, this modernized regime officially established a <strong>“SARAL” (Simple, Accessible, Rational, and Actionable)</strong> architecture. It explicitly shifts India away from an outdated, reactive model toward a proactive, privacy-by-design framework.</p><p><strong>1.1 The Pre-DPDP Regime- Section 43A and the SPDI Rules, 2011</strong></p><p>For over a decade, India’s primary data privacy protections were anchored inside <strong>Section 43A</strong> of the <strong>Information Technology (IT) Act, 2000</strong> and its subordinate legislation, the <strong>Information Technology (SPDI) Rules, 2011</strong>. This historical setup possessed several structural weaknesses that rendered it ineffective for the modern internet era:</p><ul><li><strong>Narrow “Body Corporate” Applicability-</strong> Section 43A applied strictly to commercial companies, partnerships, or sole proprietorships (“body corporates”) engaged in professional activities. It completely excluded government databases, public sector undertakings, state instrumentalities, and unincorporated individuals.</li><li><strong>Segmented Data Coverage- </strong>the 2011 Rules only protected a specific list of eight <strong>“Sensitive Personal Data or Information (SPDI)”</strong> categories such as passwords, financial data, health records, and biometrics. General personal data (e.g., names, locations, email addresses, shopping logs, metadata) fell outside this strict protection.</li><li><strong>Reactive Compensation Model- </strong>the old system focused on granting civil damages to individuals who suffered “wrongful loss or wrongful gain” due to corporate negligence. It lacked an active, regulatory enforcement body to audit architectures or penalize non-compliance.</li><li><strong>Lax Ingestion Workflows- </strong>implied consent was widely accepted. Organizations could collect data through passive methods such as pre-ticked checkmarks or continuing to use a website after seeing a privacy policy without providing itemized notices or easy ways to opt out.</li></ul><p><strong>1.2 How the DPDP Act Supersedes the Legacy System</strong></p><p>The DPDP Act completely changes India’s data governance structure. It resolves conflicts with older regulations by establishing statutory supremacy. <strong>Section 38(2)</strong> explicitly states that if any conflict arises between the DPDP Act and an existing law, the provisions of the DPDP Act prevail.</p><p>The table below outlines the core architectural and legal shifts from the old regime to the new framework:</p><figure><img alt="How the DPDP Act Supersedes the Legacy System — Pre-DPDPD vs Post-DPDP Act" src="https://cdn-images-1.medium.com/max/783/1*TFIKJMXsdMPpz0201--gXA.png" /><figcaption>How the DPDP Act Supersedes the Legacy System — Pre-DPDPD vs Post-DPDP Act | Source: Gemini</figcaption></figure><p>To clear up any legal ambiguities and avoid conflicting mandates, the DPDP Act formally alters or removes competing laws on its effective date:</p><ul><li><strong>Omission of Section 43A (IT Act)- </strong>the DPDP Act completely deletes Section 43A from the IT Act, eliminating the old “body corporate” liability framework.</li><li><strong>Decommissioning the SPDI Rules — </strong>it deletes clause (ob) under Section 87(2) of the IT Act, the exact statutory clause that allowed the government to run the SPDI Rules, 2011. This formally retires the 2011 rules.</li><li><strong>Amending the Right to Information (RTI) Act, 2005</strong> — previously, Section 8(1)(j) of the RTI Act allowed public authorities to disclose personal information if it showed a larger public interest. The DPDP Act changes this to a strict, blanket exemption — “information which relates to personal information” cannot be shared under public disclosure requests. This seals citizen data held by public offices and establishes privacy as a non-negotiable right.</li></ul><p>Furthermore, the framework’s extra-territorial mandate means that any cross-border enterprise targeting consumers within Indian territory or evaluating local behavioral trends must comply fully with these provisions.</p><h3>2. Stakeholders, Entitlements, and Civic Duties — The Human Element</h3><p>The architecture of the framework divides the digital data economy into <strong>three primary entities</strong> while pairing individual user privileges with explicit social responsibilities.</p><p>To understand how the DPDP framework functions in daily life, let us follow a single, universal use case that applies to every digital citizen- <strong><em>booking a weekly grocery delivery on “ApnaBazaar,” a popular national smartphone application. We will use “Rahul”, a regular citizen using this app, to see how the law assigns roles, protects user rights, and enforces civic responsibilities</em></strong>.</p><p><strong>2.1 Core Operational Roles (The ApnaBazaar Use Case)</strong></p><p>The framework divides the digital economy into three distinct roles based on <strong>who owns, handles, and directs the flow of data</strong>:</p><ul><li><strong>The Data Principal (The Citizen)- </strong>this is <strong>Rahul</strong>, the individual whose personal details (name, phone number, delivery address, and payment history) are being collected.<br> — <em>The Parental Clause</em><strong>- </strong>if Rahul uses the app to buy baby formula for his 7-year-old daughter, Aanya, he legally acts as the Data Principal on her behalf, because the law classifies anyone under eighteen (18) as a child requiring verified parental consent.</li><li><strong>The Data Fiduciary (The Platform)- </strong>this is <strong>ApnaBazaar LLC</strong>, the company that owns the platform. They are the data “trustees” because they determine why and how Rahul’s personal information is processed (e.g., deciding to track his location to calculate delivery times). They bear the ultimate legal liability for keeping his data safe.</li><li><strong>The Data Processor (The Third-Party Vendor)- </strong>this is <strong>“SpeedyDelivery”</strong>, an external logistics company hired by ApnaBazaar to ship groceries to Rahul’s house. SpeedyDelivery does not own Rahul’s data or decide how to use it. They receive a temporary data stream strictly to print his shipping label and navigate to his home.</li></ul><p>The diagram below illustrates the various stakeholders and defines their respective roles under the DPDP Act:</p><figure><img alt="Various Stakeholders and defines their respectiveRoles under the DPDP Act" src="https://cdn-images-1.medium.com/max/1024/1*YMfnvg1gDUn4yPuua-BHoQ.png" /><figcaption>Various Stakeholders and defines their respectiveRoles under the DPDP Act | Source: Gemini</figcaption></figure><p><strong>2.1 Enhanced User Rights (Rahul’s Entitlements)</strong></p><p>The framework gives Rahul total visibility and control over his digital footprint. Under the 2025 Rules, ApnaBazaar must fulfill any of the following requests within a maximum window of 90 days:</p><ul><li><strong>Right to Access Information- </strong>Rahul can click a button in his profile settings to request a clear, understandable summary of all the personal data ApnaBazaar has gathered about him, along with a list of all third-party processors (like SpeedyDelivery) that have handled his home address.</li><li><strong>Right to Correction and Updatin- </strong>if Rahul notices that a typo in his profile reads “Rauhyl” instead of “Rahul,” or if he moves to a new apartment, he has the legal right to compel ApnaBazaar to fix the inaccuracy and update his records instantly across all their live servers .</li><li><strong>Right to Erasure (“The Right to be Forgotten”)- </strong>if Rahul decides to delete his ApnaBazaar account, he can demand that the platform wipe his files. ApnaBazaar must completely delete his data from its live production databases and cloud storage clusters.<br> — <em>The Legal Exception</em><strong>- </strong>if tax or banking laws require ApnaBazaar to keep invoices of Rahul’s past financial transactions for audit purposes, they may safely retain only those specific records.</li><li><strong>Right to Nominate- </strong>in his account privacy dashboard, Rahul can designate his brother, Amit, as his official proxy. If Rahul experiences a severe medical emergency that leaves him incapacitated, or if he passes away, Amit gains the legal right to manage or permanently delete Rahul’s ApnaBazaar account profile.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aErZMLa7aZs3PqBaR8Q0gA.png" /><figcaption>Enhanced User Rights (Rahul’s Entitlements) — Rahul’e My Profile View | Source: Gemini</figcaption></figure><p><strong>2.2 Statutory Civic Duties (Rahul’s Responsibilities)</strong></p><p>To maintain a fair ecosystem, the framework establishes that privacy is a two-way street. Rahul is legally bound to several civic responsibilities while using the application:</p><ul><li><strong>The Duty of Authenticity- </strong>Rahul cannot create an account under the name “Chief of Police Sharma” or use an identity that isn’t his to get special platform privileges, which means impersonation is strictly forbidden.</li><li><strong>No Material Suppression- </strong>if Rahul submits official identity documents to access a restricted or subsidized grocery marketplace option run by a government initiative, he cannot hide or alter any material background information.</li><li><strong>Protection Against Frivolous Claims- </strong>if an ApnaBazaar delivery is simply late, Rahul cannot file a false, malicious data breach complaint with the <strong>Data Protection Board</strong> out of anger. The law prohibits filing frivolous grievances. If Rahul weaponizes the regulatory system in bad faith, he can face administrative penalties up to ₹10,000.</li></ul><p>This framework establishes clearly defined boundaries to protect privacy, ensuring that routine tasks such as online grocery shopping stay secure, transparent, and under the citizen’s control. This represents merely one of the countless ways the DPDP Act impacts our daily lives.</p><p>The diagram below depicts the overarching high-level regulatory framework established under the DPDP Act:</p><figure><img alt="Overarching Regulatory Framework for DPDP Act" src="https://cdn-images-1.medium.com/max/1024/1*Uw1Uxu1eDaw111wFxk9BAQ.png" /><figcaption>Overarching Regulatory Framework for DPDP Act | Source: Gemini</figcaption></figure><p><strong>Quick Reference: Who Does What?</strong></p><p>Based on the previously discussed use case, the illustration below outlines the specific roles and actions of each stakeholder under the DPDP Act:</p><figure><img alt="DPDP Act Stakeholder Roles (ApnaBazaar use case)" src="https://cdn-images-1.medium.com/max/679/1*O_Ek_h404smIYvLwjibfgw.png" /><figcaption>DPDP Act Stakeholder Roles (ApnaBazaar use case) | Source: Gemini</figcaption></figure><h3>3. Core Governance and Operational Workflows — The User Interface Transformation</h3><p>Transitioning an organization to any compliance requires transforming how data is ingested, managed, and reported.</p><p>When Rahul opens the updated ApnaBazaar app to sign up for his grocery account, he should no longer see a single, tiny checkbox that says, <strong>“I agree to the unreadable 45-page Terms of Service and Privacy Policy.”</strong> Under the DPDP framework, that legacy method of extracting “bundled” or implied consent is entirely illegal.</p><p>To comply with the law, ApnaBazaar must structurally alter its user registration screens and background data pipelines.</p><p><strong>3.1 UI/ UX Mapping: The Compliant Sign-Up Screen</strong></p><p>Every consent request initiated by ApnaBazaar must be preceded or accompanied by an explicit, itemized privacy notice. This notice must state clearly what exact types of data are being collected and the distinct purpose behind processing them. It must also detail exactly how Rahul can exercise his rights to revoke consent or edit his profile records, and provide explicit paths for Rahul to file complaints with the Data Protection Board of India.</p><p>The interface must present this choice in a clear and plain language format , granting Rahul the explicit choice to access the notice in English or any of the <strong>22 regional languages</strong> listed under the <strong>Eighth Schedule of the Constitution</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Doz9_Awj9JHiHpTjUzKbSA.png" /><figcaption>The Blueprint of ApnaBazaar’s New Compliant-ready Registration Screen | Source: Gemini</figcaption></figure><p><strong>3.2 Behind-the-Screen Operational Workflows</strong></p><p>To support this new layout, ApnaBazaar’s engineering teams must operationalize three core background workflows:</p><p><strong><em>Workflow A: The “Comparable Ease” Erasure Pipeline</em></strong></p><p>If Rahul decides to toggle <strong>“Live GPS Location Tracking”</strong> to <strong>[OFF]</strong> three weeks after signing up, the background application code must react immediately. The framework requires that withdrawing consent must be just as easy as giving it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WKTlWDOTpceJ_t_T-VErGQ.png" /><figcaption>Workflow A: The “Comparable Ease” Erasure Pipeline | Source: Gemini</figcaption></figure><p>ApnaBazaar must instantly cease tracking his coordinates and command its third-party logistics processor, SpeedyDelivery, to clear Rahul’s location telemetry from its active servers.</p><p><strong><em>Workflow B: The 72-Hour Emergency Response Protocol</em></strong></p><p>What happens if an unauthorized third party cracks an unsecured API database endpoint at ApnaBazaar, exposing Rahul’s home address and password hash? Under the old regime, companies could quietly patch leaks without notifying customers. Under the DPDP framework, a data breach requires immediate, systemic action:</p><ol><li><strong>Internal Identification- </strong>ApnaBazaar’s monitoring tools flag an anomalous data export. The security team confirms a personal data breach has occurred.</li><li><strong>The 72-Hour Regulatory Filing- </strong>within 72 hours of identifying the incident, ApnaBazaar is legally required to submit an official, comprehensive report to the Data Protection Board of India. This report must detail the cause, the volume of affected users, and immediate technical mitigation steps taken.</li><li><strong>Direct User Warning- </strong>simultaneously, ApnaBazaar must push a clear notification to Rahul (via email, SMS, or in-app message) warning him of the compromise and advising him to reset his security credentials immediately.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3FQg4NDAnTyDdycicuP_ZA.png" /><figcaption>Workflow B: The 72-Hour Emergency Response Protocol | Source: Gemini</figcaption></figure><p><strong><em>Workflow C: The Phased Transition Integration</em></strong></p><p>ApnaBazaar cannot fix its entire legacy architecture overnight. The 2025 Rules provide a structured <strong>18-month rollout runway</strong> to map its current code bases:</p><ul><li><strong>Legacy Notice Updates- </strong>for users who signed up before the Act took effect, ApnaBazaar must dispatch a retrospective digital notice via email or push alert as soon as reasonably practicable. This notice must explicitly describe what historical data is currently held and remind users of their rights to delete or modify it.</li><li><strong>Consent Manager Integration- </strong>by the 12-month mark, ApnaBazaar’s technical teams must build open, interoperable API endpoints. These endpoints will allow Rahul to review, track, and pull back his permissions through an external, registered centralized <strong>“Consent Manager”</strong> app, independent of ApnaBazaar’s native settings.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XulzoQy8lLu8MOabhpZ0XQ.png" /><figcaption>Workflow C: The Phased Transition Integration : Source: Gemini</figcaption></figure><p><strong>3.3 Core UI Checklist for Data Fiduciaries</strong></p><p>Organizations can evaluate their registration interfaces against this scannable compliance metric:</p><ol><li><strong>Granular Options- </strong>are non-essential data captures (like location tracking or marketing profiles) unbundled and set to “OFF” by default?</li><li><strong>Language Selection- </strong>can the user change the entire notice text into any of the 22 official Indian regional languages with a single click?</li><li><strong>Direct Contact Data- </strong>does the registration screen explicitly list the corporate email or contact path for the Data Protection Officer?</li><li>[ ] <strong>Clear Exit Path- </strong>is the button to revoke consent just as easy to find and click as the button used to accept the terms?</li></ol><figure><img alt="Core UI Checklist for Data Fiduciaries" src="https://cdn-images-1.medium.com/max/1024/1*s5V7Kwj4TX5N4IgUIUqnWg.png" /><figcaption>Core UI Checklist for Data Fiduciaries : Source: Gemini</figcaption></figure><h3>4. Technology Architecture and Security Requirements</h3><p>Implementing Technology Architecture and Security Requirements fundamentally means <strong>converting written policies into executable code</strong>. Currently, numerous organizations face confusion and difficulties in identifying and mapping the right technologies needed to transition from policy text to operational solutions within stipulated timelines.</p><p>For ApnaBazaar’s engineering teams, adhering to the DPDP framework goes well beyond revising a privacy policy, it necessitates a complete re-engineering of data isolation, user authentication protocols, and automated data destruction mechanisms.</p><p>Here is the high-level technical blueprint ApnaBazaar must implement to satisfy these stringent security mandates.</p><p><strong>Building Cryptographic Perimeters</strong></p><p>To prevent unauthorized access and ensure data integrity, ApnaBazaar must move away from perimeter-only defence (like standard firewalls) and adopt a Zero-Trust, Data-Centric Security Model.</p><ul><li><strong>End-to-End Encryption (E2EE)/ Data-in-Transit-</strong> all data moving between Rahul’s mobile app, ApnaBazaar’s web servers, and third-party processors (like SpeedyDelivery) must use TLS 1.3.<br> — <em>Data-at-Rest-</em> personal identifiers (names, phone numbers, GPS coordinates) must be encrypted using AES-256 at the database level. Field-level encryption should be applied to highly sensitive fields so that even if a database is compromised, the data remains unreadable.</li><li><strong>Enclave-Based Processing &amp; Tokenization- i</strong>nstead of passing raw user data across internal microservices (e.g., sending Rahul’s real phone number to the analytics engine), ApnaBazaar should use a tokenization vault. The real data is substituted with a reversible cryptographic token.</li><li><strong>Robust Key Management- c</strong>ryptographic keys must be segregated from the data they protect. Engineers should utilize a cloud-based Hardware Security Module (HSM) with automated key rotation policies.</li></ul><p><strong>Configuring Age-Gating for Kids</strong></p><p>The DPDP Act places a massive premium on protecting children’s data (defined as individuals under 18). ApnaBazaar cannot profile children or target advertisements at them.</p><p><strong>The Age-Verification Gate</strong></p><p>ApnaBazaar must implement a frictionless but legally defensible mechanism to determine user age during registration or profile updates:</p><ul><li><strong>Agnostic Verification- u</strong>sing zero-knowledge proofs (ZKP) or tokenized third-party identity verification providers to confirm age without storing the user’s actual government ID documents on ApnaBazaar’s servers.</li><li><strong>The Parental Consent Fork- i</strong>f the system detects the user is a minor, the workflow must immediately redirect to capture Verifiable Parental Consent (VPC).</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kors1994AP5G-0MmdY3VOQ.png" /><figcaption>Verifiable Parental Consent (VPC) | Source: Gemini</figcaption></figure><ul><li><strong>Database Tagging- o</strong>nce a profile is flagged as a minor, a strict database flag (is_minor = TRUE) must automatically trigger an API-level block on all behavioral tracking scripts, push notifications containing marketing, and data-sharing pipelines with non-essential third parties.</li></ul><p><strong>Managing Automated Data Deletion Routines</strong></p><p>Data cannot sit in ApnaBazaar’s databases indefinitely. When Rahul withdraws his consent, or when his order is successfully delivered and the business purpose is fulfilled, his data must be systematically purged.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nGIR7ZS6Utc-lSUVy_gcQw.png" /><figcaption>Automating and Optimizing Data Purging for Compliance | Source: Gemini</figcaption></figure><p>Engineers must build an <strong>Automated Purge Engine</strong> that operates across the entire technical stack.</p><ul><li><strong>Time-to-Live (TTL) Configurations- f</strong>or temporary data (like live GPS tracking during a delivery), strict database TTLs must be set. The moment SpeedyDelivery marks an order as “Delivered,” a countdown timer triggers, wiping the spatial telemetry within a strictly defined window (e.g., 48 hours for dispute resolution).</li><li><strong>Cascading Deletions via Orchestration- d</strong>eleting data from the primary database is not enough. ApnaBazaar’s purge engine must use webhook listeners to broadcast a “Delete Request” across all environments:<br> — <em>Production DBs</em>- hard delete or irreversible anonymization of Rahul’s user row.<br> — <em>Data Warehouses/Logs- s</em>crubbing analytical pipelines and cold storage logs using automated regex pruning scripts.<br> — <em>Downstream Processors- t</em>riggering an API call to SpeedyDelivery’s endpoints forcing them to delete Rahul’s address from their systems.</li></ul><p><strong>The “Right to be Forgotten” Challenge</strong></p><p>A major headache for the engineering team will be backup retention. If Rahul requests erasure, completely rewriting multi-terabyte immutable backup tapes instantly is practically impossible.</p><p>The cryptographic vault, the parental verification loop, and the cascading deletion engine pose the highest technical debt or integration challenge for organizations existing legacy infrastructure.</p><p><strong><em>One of many potential Technical Solution</em></strong>- ApnaBazaar must maintain an encrypted “Deletion Ledger.” When backups are restored during a disaster recovery event, the system must immediately parse this ledger and re-apply the deletions before the environment goes live.</p><h3>5. Enforcement Metrics &amp; Financial Penalty Structure</h3><p>The framework officially removes previous loose liability laws (completely omitting Section 43A of the Information Technology Act, 2000) and replaces them with an aggressive, non-compliance financial structure.</p><p>Administrative penalties are scaled based on the repetition, scale, duration, and financial impact of the infraction.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p_Pp2edfGtjTIYXHSp8GMg.png" /><figcaption>Enforcement Metrics &amp; Financial Penalty Structure | Source: Gemini</figcaption></figure><p><strong>Appellate Review Window</strong></p><p>Any business or individual wishing to challenge a formal directive or financial sanction issued by the Data Protection Board can file an appeal before the <strong>Appellate Tribunal (TDSAT)</strong> within a maximum period of 60 days from receiving the order.</p><h3>6. Summary — The Path Forward</h3><p>Transitioning an enterprise like ApnaBazaar LLC into a DPDP-compliant organization is not a one-time legal exercise. It is a fundamental restructuring of engineering culture, data strategy, and user trust.</p><p>By mapping out Rahul’s journey from a simple consumer to an empowered Data Principal, we can see that compliance is a living ecosystem. When Rahul flips a location toggle to [OFF], it isn’t just a UI change, it triggers a cascade of cryptographic events, database purges, and third-party processor synchronizations with SpeedyDelivery.</p><p><strong>The Executive Roadmap</strong></p><p>To successfully navigate the 18-month runway provided by the framework, an enterprise’s leadership and engineering teams must coordinate across three main horizons:</p><figure><img alt="The Executive Roadmap for DPDP Compliance" src="https://cdn-images-1.medium.com/max/1024/1*vgBUeQRBTsa5GDeglQluJA.png" /><figcaption>The Executive Roadmap for DPDP Compliance | Source: Gemini</figcaption></figure><p><strong>The Core Mindset Shift</strong></p><p>Compliance shouldn’t be viewed as a technical tax or a roadblock to innovation. In the modern digital economy, privacy is a product feature. The guardrails built today to protect users like Rahul are the exact foundations that will prevent catastrophic data liabilities and build deep, unbreakable brand loyalty tomorrow.</p><p>Having examined the fundamentals of the DPDP Act and its impact on daily life, we have detailed the comprehensive stakeholder map, emergency breach protocols, legacy transition timelines, and technical architecture. If you would like to understand how these requirements align with your specific enterprise and industry, and begin a readiness assessment to evaluate your current posture, please contact us via <a href="/service/mailto:send2joy@gmail.com">send2joy@gmail.com</a> | <a href="/service/mailto:joy@wso2.com">joy@wso2.com</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dyXRCqOevzqeDH2_mbF4Xg.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=78b9c99d7ab6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Build Power Automate Flows Faster with AI]]></title>
            <link>https://jrathnayake.medium.com/build-power-automate-flows-faster-with-ai-5810cee0a919?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/5810cee0a919</guid>
            <category><![CDATA[power-automate]]></category>
            <category><![CDATA[powerplatform]]></category>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[azure-ai]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Fri, 03 Oct 2025 04:55:33 GMT</pubDate>
            <atom:updated>2025-10-03T04:55:33.664Z</atom:updated>
            <content:encoded><![CDATA[<p>Artificial Intelligence (AI) has permeated nearly every aspect of our technological landscape, and its influence is now deeply integrated into the Microsoft Power Platform.</p><p>This video will delve into the exciting ways AI can accelerate and enhance the development of Power Automate Flows, empowering users to create more efficient and sophisticated automation solutions with unprecedented speed. We will explore various AI capabilities and how they can be leveraged to streamline the design, implementation, and optimization of your workflows.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FBchh3l5KRZI%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DBchh3l5KRZI&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FBchh3l5KRZI%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/68ccd257b33e7e8a5b45bee4027817eb/href">https://medium.com/media/68ccd257b33e7e8a5b45bee4027817eb/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zl32k6K_wGfbTYRsNYi6nw.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5810cee0a919" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Getting Started with Azure OpenAI Service]]></title>
            <link>https://jrathnayake.medium.com/getting-started-with-azure-openai-service-1977e640bb1e?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/1977e640bb1e</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[azureopenai]]></category>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[azure-ai]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Tue, 30 Sep 2025 04:46:35 GMT</pubDate>
            <atom:updated>2025-09-30T04:46:35.872Z</atom:updated>
            <content:encoded><![CDATA[<p>The <strong>Azure OpenAI Service</strong> empowers developers to integrate the cutting-edge capabilities of generative Artificial Intelligence directly into their applications. This powerful service allows you to leverage advanced language models, such as <strong>GPT-4</strong>, to create intelligent, conversational, and highly interactive solutions.</p><p>Through the intuitive <strong>Azure portal</strong>, you can seamlessly manage the entire lifecycle of your AI integration. This includes:</p><ul><li><strong>Setting Up Your Environment:</strong> Easily configure your Azure environment to host and utilize the OpenAI Service.</li><li><strong>Deploying Language Models:</strong> Deploy pre-trained, powerful language models like GPT-4 with just a few clicks, making them accessible to your applications.</li><li><strong>Integrating Conversational AI:</strong> Incorporate sophisticated conversational AI features, enabling your applications to understand and generate human-like text, respond to queries, and engage in dynamic dialogues.</li></ul><p>The Azure OpenAI Service is an invaluable resource for <strong>developers exploring cutting-edge AI tools</strong>. It provides a robust, scalable, and secure platform to unlock the full potential of generative AI, facilitating the creation of innovative applications across various industries and use cases.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FEfAEOL_YaAE%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DEfAEOL_YaAE&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FEfAEOL_YaAE%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/70d7b2202480a3cc7623fa502d801b26/href">https://medium.com/media/70d7b2202480a3cc7623fa502d801b26/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pSnZNDt93BPxQh3Ldc1Tiw.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1977e640bb1e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Software Development Kit (SDK) support in WSO2 API Manager]]></title>
            <link>https://jrathnayake.medium.com/software-development-kit-sdk-support-in-wso2-api-manager-90995c1c8f0d?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/90995c1c8f0d</guid>
            <category><![CDATA[wso2-api-manager]]></category>
            <category><![CDATA[api-management]]></category>
            <category><![CDATA[sdk]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[wso2]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Fri, 12 Sep 2025 07:29:43 GMT</pubDate>
            <atom:updated>2025-09-12T07:29:43.849Z</atom:updated>
            <content:encoded><![CDATA[<p>API consumers efficiently use client-side SDKs to build applications that interact with subscribed APIs. These SDKs abstract complexities like HTTP requests and authentication, accelerating development and improving code quality through strongly typed interfaces and best practices.</p><p>WSO2 API Manager facilitates this by allowing API providers to expose APIs for easy SDK generation and integration, fostering a thriving API ecosystem.</p><p>Enabling additional SDKs within WSO2 API Manager involves understanding its architecture and configuration options, expanding supported languages and frameworks, and integrating with code generation tools to simplify API consumption for diverse developer audiences.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FD27uiaX4OUo%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DD27uiaX4OUo&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FD27uiaX4OUo%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/67c73c7e13047e0317303d9951ef4691/href">https://medium.com/media/67c73c7e13047e0317303d9951ef4691/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YTzzQXX7vN_fgaZl1UyV-g.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=90995c1c8f0d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Getting Started with Azure AI Computer Vision Service]]></title>
            <link>https://jrathnayake.medium.com/getting-started-with-azure-ai-computer-vision-service-01a1b9310140?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/01a1b9310140</guid>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[azure-computer-vision]]></category>
            <category><![CDATA[azure-ai]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Fri, 05 Sep 2025 04:10:03 GMT</pubDate>
            <atom:updated>2025-09-05T04:11:14.288Z</atom:updated>
            <content:encoded><![CDATA[<p>The Azure AI Computer Vision Service offers a comprehensive suite of powerful capabilities designed to enable developers to integrate advanced image and video analysis into their applications. By leveraging these features, businesses and individuals can unlock new insights from visual data, automate complex tasks, and create more engaging user experiences.</p><p>One of the core functionalities of the service is <strong>image analysis</strong>, which allows for the extraction of various visual features from images. This includes, but is not limited to:</p><ul><li><strong>Object Detection:</strong> Identifying and locating specific objects within an image, such as people, vehicles, or animals, and providing bounding box coordinates. This is crucial for applications like inventory management, security surveillance, and autonomous systems.</li><li><strong>Facial Recognition:</strong> Detecting human faces, identifying facial landmarks, and even recognizing individuals based on a pre-trained model. This capability is vital for applications requiring user authentication, personalized experiences, and demographic analysis.</li><li><strong>Optical Character Recognition (OCR):</strong> Extracting printed or handwritten text from images, converting it into machine-readable format. This is invaluable for digitizing documents, processing forms, and enabling searchable content from visual sources.</li><li><strong>Attribute Recognition:</strong> Identifying attributes like age, gender, and emotional state of individuals in an image, or colour schemes and dominant colours in a scene.</li><li><strong>Image Tagging:</strong> Generating a list of relevant tags or keywords that describe the content of an image, making it easier to categorize, search, and manage large image collections.</li><li><strong>Content Moderation:</strong> Detecting and flagging potentially inappropriate or harmful content in images, helping to maintain a safe and compliant digital environment.</li></ul><p>Beyond static image analysis, the Azure AI Computer Vision Service also provides capabilities for <strong>video processing</strong>, enabling real-time or batch analysis of video streams. This can involve:</p><ul><li><strong>Motion Detection:</strong> Identifying movement within video frames.</li><li><strong>Activity Recognition:</strong> Recognizing specific actions or behaviors occurring in a video.</li><li><strong>Scene Understanding:</strong> Analyzing the overall context and events unfolding in a video.</li></ul><p>To leverage these capabilities in your applications, the Azure AI Computer Vision Service offers a set of user-friendly <strong>APIs (Application Programming Interfaces)</strong> and <strong>SDKs (Software Development Kits)</strong> that support various programming languages. Developers can integrate the service by making API calls, sending image or video data, and receiving structured JSON responses containing the analysis results.</p><p>The applications of the Azure AI Computer Vision Service are vast and constantly expanding, encompassing areas such as:</p><ul><li><strong>Retail:</strong> Product recognition, shelf analytics, personalized recommendations.</li><li><strong>Healthcare:</strong> Medical image analysis, diagnostic assistance.</li><li><strong>Manufacturing:</strong> Quality control, anomaly detection.</li><li><strong>Security:</strong> Surveillance, access control, threat detection.</li><li><strong>Education:</strong> Content categorization, learning material analysis.</li><li><strong>Media and Entertainment:</strong> Content moderation, metadata generation, visual search.</li></ul><p>By embracing the power of the Azure AI Computer Vision Service, developers can build intelligent applications that understand and interact with the visual world, driving innovation and efficiency across various industries.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FtiOdKXEa0WQ%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DtiOdKXEa0WQ&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FtiOdKXEa0WQ%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/16f2e0e72525f1f5d86a4b82d09fe112/href">https://medium.com/media/16f2e0e72525f1f5d86a4b82d09fe112/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vd2j85mz-m8XJidZ.png" /><figcaption>Azure AI Computer Vision Services</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=01a1b9310140" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Getting Started with Azure AI Services]]></title>
            <link>https://jrathnayake.medium.com/getting-started-with-azure-ai-services-4af4f1769933?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/4af4f1769933</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[azure-ai]]></category>
            <category><![CDATA[azure]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Fri, 05 Sep 2025 03:52:54 GMT</pubDate>
            <atom:updated>2025-09-05T03:52:54.892Z</atom:updated>
            <content:encoded><![CDATA[<p>Microsoft Azure AI Services represents a rapidly evolving suite of Software-as-a-Service (SaaS) artificial intelligence capabilities.</p><p>This video presentation aims to provide a comprehensive exploration of what Azure AI Services encompass, delving into their diverse functionalities and the rich array of resources that are readily available to application developers. We will examine how these services empower developers to integrate advanced AI features into their applications with ease, accelerating innovation and enhancing user experiences.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FfPkVh4NgO0M%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DfPkVh4NgO0M&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FfPkVh4NgO0M%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/321b52b1bd4302b61db0822af699238d/href">https://medium.com/media/321b52b1bd4302b61db0822af699238d/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*uwXAf2WfTf0Cj8J9.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4af4f1769933" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Developing integrations became super super easy - Introducing VSCode extension for WSO2 Micro…]]></title>
            <link>https://jrathnayake.medium.com/wso2-is-excited-to-announce-the-release-of-a-new-vscode-extension-specifically-designed-for-the-b27e6cf6f3d6?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/b27e6cf6f3d6</guid>
            <category><![CDATA[enterprise-integration]]></category>
            <category><![CDATA[wso2]]></category>
            <category><![CDATA[intergration]]></category>
            <category><![CDATA[wso2-micro-integrator]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Tue, 02 Sep 2025 07:39:35 GMT</pubDate>
            <atom:updated>2025-09-02T07:40:28.847Z</atom:updated>
            <content:encoded><![CDATA[<p>WSO2 is excited to announce the release of a new VSCode extension specifically designed for the WSO2 Micro Integrator. This innovative tool significantly simplifies the development of integration solutions, particularly benefiting citizen developers who may not have extensive coding experience.</p><p>The VSCode extension for WSO2 Micro Integrator empowers users with a streamlined and intuitive environment to create, test, and deploy integration flows. Here’s a closer look at its key features and how they enhance the development experience:</p><p><strong>Key Features of the VSCode Extension for WSO2 Micro Integrator:</strong></p><ul><li><strong>Visual Flow Editor:</strong> The extension provides a drag-and-drop interface for designing integration sequences. Users can visually connect various endpoints, mediators, and connectors, making it easy to understand the flow of data and logic. This visual approach eliminates the need for manual XML configuration, accelerating development and reducing errors.</li><li><strong>Built-in Connectors and Mediators:</strong> A rich library of pre-built connectors for popular applications and services (e.g., Salesforce, SAP, databases, message queues) is readily available. Additionally, a comprehensive set of mediators allows for common data transformations, routing logic, error handling, and more. This extensive library reduces development time by providing ready-to-use components.</li><li><strong>Live Debugging Capabilities:</strong> Developers can easily set breakpoints and step through their integration flows, inspect variable values, and monitor message payloads in real-time. This robust debugging functionality significantly aids in identifying and resolving issues quickly, leading to more reliable integrations.</li><li><strong>Schema Validation and Auto-Completion:</strong> The extension offers intelligent assistance with schema validation, helping users ensure that their integration configurations adhere to the correct structure and syntax. Auto-completion suggestions for various elements and attributes further enhance the development speed and accuracy.</li><li><strong>Unit Testing Framework:</strong> The extension integrates with a unit testing framework, enabling developers to create and run tests for their integration flows. This ensures that integrations function as expected and helps maintain code quality.</li><li><strong>Deployment Integration:</strong> Users can directly deploy their developed integration artifacts to a running WSO2 Micro Integrator instance from within VSCode. This seamless deployment process simplifies the transition from development to production.</li><li><strong>Version Control Integration:</strong> Standard VSCode features for Git and other version control systems are fully supported, allowing teams to collaborate effectively on integration projects, track changes, and manage different versions of their integration flows.</li><li><strong>Developer-Friendly Interface:</strong> The extension is designed with a focus on usability, offering a clean and organized interface that is easy to navigate, even for users new to integration development.</li></ul><p>By providing these powerful features within the familiar and popular VSCode environment, the WSO2 Micro Integrator extension significantly lowers the barrier to entry for integration development. Citizen developers, business analysts, and even seasoned integration specialists can now efficiently build, deploy, and manage robust integration solutions, accelerating digital transformation initiatives within their organizations. This extension represents a significant step towards democratizing integration development, empowering a broader range of users to contribute to connected enterprise architectures.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FgMbHgGkMe5I%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DgMbHgGkMe5I&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FgMbHgGkMe5I%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/75a5b5c2d70c9709cfe505bbd1ee25f5/href">https://medium.com/media/75a5b5c2d70c9709cfe505bbd1ee25f5/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*2h-UVuaaER5l3kN7.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b27e6cf6f3d6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Using Input Parameters and Expressions in Power Automate]]></title>
            <link>https://jrathnayake.medium.com/using-input-parameters-and-expressions-in-power-automate-201725770f6e?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/201725770f6e</guid>
            <category><![CDATA[microsoft-power-platform]]></category>
            <category><![CDATA[microsoft]]></category>
            <category><![CDATA[power-automate]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Tue, 02 Sep 2025 07:14:00 GMT</pubDate>
            <atom:updated>2025-09-02T07:14:00.984Z</atom:updated>
            <content:encoded><![CDATA[<p>I recently performed a detailed screen capture to illustrate a crucial aspect of our system’s functionality: the ability to efficiently capture and process various input parameters when initiating a workflow using Power Automate.</p><p>This demonstration also highlighted the versatility and power of our expression framework, showcasing how it can be leveraged to effectively manipulate data and information within a given flow. The objective was to provide a clear, practical guide for users to understand how seamlessly they can define and utilize inputs, and subsequently transform or extract valuable insights from the data using our robust set of built-in expressions available in Power Automate. This capability is fundamental to building dynamic and adaptable workflows that can respond to diverse user needs and data structures.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Fo3h6zEe-Hwk%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Do3h6zEe-Hwk&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fo3h6zEe-Hwk%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/59b6e4d2e8f78cf69e1aaeb0d9c276b8/href">https://medium.com/media/59b6e4d2e8f78cf69e1aaeb0d9c276b8/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=201725770f6e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Open-Dev Summit 2025 — UoK]]></title>
            <link>https://jrathnayake.medium.com/open-dev-summit-2025-uok-49fa9071cfe2?source=rss-50cfe99c97f3------2</link>
            <guid isPermaLink="false">https://medium.com/p/49fa9071cfe2</guid>
            <category><![CDATA[identity-management]]></category>
            <category><![CDATA[identity]]></category>
            <dc:creator><![CDATA[Joy Rathnayake]]></dc:creator>
            <pubDate>Tue, 02 Sep 2025 06:34:02 GMT</pubDate>
            <atom:updated>2025-09-02T06:34:02.249Z</atom:updated>
            <content:encoded><![CDATA[<h3>Open-Dev Summit 2025 — UoK</h3><p>The Open-Dev Summit 2025, a significant event orchestrated by the University of Kelaniya, was conceived with a clear and crucial objective: to effectively bridge the existing gap between theoretical academic learning and the practical, dynamic demands of the industry. This ambitious initiative was designed to provide participants with a comprehensive and immersive experience, thereby enhancing their readiness for the professional world.</p><p>A pivotal component of this overarching initiative was the inclusion of five highly impactful pre-summit workshops. These workshops were strategically crafted to offer participants hands-on experience and in-depth knowledge in various cutting-edge technologies and industry best practices, setting a robust foundation for the main summit.</p><p>As an invited speaker at their most recent summit, organized by the proactive and enthusiastic FOSS (Free and Open Source Software) Community of the University of Kelaniya, I had the privilege of delivering a session. During my participation, I observed a remarkable synergy of energy, a spirit of collaborative learning, and, most notably, a genuine passion for acquiring knowledge among all the participants. This collective enthusiasm underscored the success of the summit in fostering an engaging environment for both learning and networking.</p><p>I led a session on “Modernizing Authentication and Authorization with IAM”, presenting real-world use cases and elaborating on some industry best practices.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Cye07GmsmR7RzJUakgyynw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dCsbGb1R7YwFWm4fyQXk_A.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HGUEdoZP8MZTgIwgPSyYTg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iyRy-Lo56D4hRxogLEst5g.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9-LLbfuqnPxjgaR_tKRXgQ.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=49fa9071cfe2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>