<?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 Bukola Eunice on Medium]]></title>
        <description><![CDATA[Stories by Bukola Eunice on Medium]]></description>
        <link>https://medium.com/@bukola-eo?source=rss-8209cfc2a06a------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*xVHmAyHzlHj3ar3rVzvSCA.jpeg</url>
            <title>Stories by Bukola Eunice on Medium</title>
            <link>https://medium.com/@bukola-eo?source=rss-8209cfc2a06a------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 15:34:14 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@bukola-eo/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[Every Millisecond Has a Price Tag — Why Latency Testing Saves (or Sinks) Fintech/SaaS]]></title>
            <link>https://bukola-eo.medium.com/every-millisecond-has-a-price-tag-why-latency-testing-saves-or-sinks-fintech-saas-fd5de30cf83f?source=rss-8209cfc2a06a------2</link>
            <guid isPermaLink="false">https://medium.com/p/fd5de30cf83f</guid>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[software-qa]]></category>
            <category><![CDATA[performance-testing]]></category>
            <category><![CDATA[qa-testing]]></category>
            <dc:creator><![CDATA[Bukola Eunice]]></dc:creator>
            <pubDate>Tue, 31 Mar 2026 21:38:36 GMT</pubDate>
            <atom:updated>2026-03-31T21:38:36.898Z</atom:updated>
            <content:encoded><![CDATA[<p>In a document editor, 500 milliseconds of lag is annoying. In a photo storage app, it’s frustrating. But in fintech or SaaS, where every screen shows balances, executes trades, or processes payments, 500ms can mean: <strong>a double charge because a user clicked twice, a failed trade because the price moved before confirmation, an overdraft because the balance shown was stale, a support ticket that costs $15 to resolve.</strong></p><p>Every SaaS and Fintech SaaS company proudly displays its uptime badge: “99.99% available over the last 90 days.” Customers nod approvingly. Investors feel warm inside, but here’s the uncomfortable truth: <strong>uptime says nothing about speed.</strong></p><p>The purpose of my writing this article is to expound to you why latency testing is not “performance tuning” in fintech/SaaS. It’s financial risk management, and skipping it is a 99.99% lie that customers will eventually discover.</p><p>What do I mean by latency testing? <strong>Latency testing</strong> is simply checking how long it takes for something to respond after you trigger it.</p><p>Think of it like this: You click a button → how many seconds (or milliseconds) before something happens?</p><p>You see that delay that often occurs; it is what we call <strong>latency.</strong></p><p>Now, let’s talk about why Latency hits fintech harder</p><p>Believe it or not, <strong>delayed data always has a domino effect.</strong> Regular SaaS has loose consistency. A comment appearing 2 seconds late? Fine. A friend request notification is delayed? Whatever. Fintech is in a financial state, and that state has a time bomb attached. Let’s consider a simple balance check:</p><p>User opens app → sees $500 balance → spends $450 But the balance was actually $400, just 3 seconds stale</p><p>Result: $50 overdraft, $35 fee, angry customer</p><p>The delay wasn’t intentionally malicious. It was just latency between the database, cache, and API. If there’s one thing I need you to understand as you read this, it’s that in Fintech or Fintech SaaS, stale data is dead data.</p><p>The second reason is a <strong>race-condition trap.</strong> Race conditions happen when two operations compete for the same resource, and latency decides the winner.</p><p>Classic example: User clicks “Cancel Subscription” twice because the first click seemed unresponsive.</p><p><em>Click #1: Begins cancellation</em></p><p><em>Click #2: Also begins cancellation (server hasn’t updated yet)</em></p><p>Result: Double refund, double webhook, angry accountant, and customer.</p><p>In regular SaaS, a duplicate cancel might just error gracefully. In fintech, it’s real money moving twice.</p><p>Thirdly, there is always an <strong>uncontrollable variable here: third-party gateways.</strong></p><p>Your API might respond in 50ms. But Stripe? PayPal? A European bank’s legacy SOAP service? They might respond in 800ms or 3 seconds, or time out entirely.</p><p>Fintech SaaS is only as fast as its slowest dependency, and we can’t control that dependency.</p><p>Let me give you <strong>real-world wounds</strong>, i.e., examples you can learn from, so you get me clearly.</p><p><strong>Case 1:</strong> A Neobank that lets users spend money twice.</p><p>A digital wallet application, let’s call it “SparkleBank,” had a lovely real-time balance display. Real-time, in this case, meant a 2-second cache.</p><p>During a flash sale, users frantically clicked “Pay.”</p><p>The application displayed a $100 balance for two users simultaneously. Both clicks succeeded, and both users spent the same $100.</p><p><strong>The root cause:</strong> balance read latency, no optimistic locking.</p><p><strong>The result:</strong> $10,000 in double-spent money, 3 days to resolve, 1 regulatory inquiry.</p><p><strong>What QA lesson is to be learned here? </strong>Balance reads under concurrent user simulation, simulating latency.</p><p><strong>Case 2: </strong>The Crypto SaaS solution</p><p>The company offered a crypto “real-time” price dashboard. However, the price updates were delayed by 800ms due to database replication issues.</p><p>The arbitrage bots caught on. They would read the outdated price, buy through the API, then sell at the updated price a few milliseconds later. The company lost $50,000 before figuring this out.</p><p><strong>What went wrong?</strong> Stale reads + lack of price staleness checks.</p><p><strong>What can we learn from this?</strong> Financial loss masquerading as a performance problem.</p><p><strong>QA takeaway?</strong> Don’t forget to check for arbitrage scenarios. Any price delay is an opportunity for bots or even humans to take over.</p><p><strong>Case 3:</strong> A promo code that charged full price</p><p>The user tries to subscribe with a 50% discount promo code during checkout. The promo validation takes a long time, 2 seconds, to respond. The user, seeing no response, decides to click on the subscribe button again.</p><p>The second click happens before they are notified of the discount, and they are charged full price.</p><p><strong>The root cause</strong> of this problem is the long time the promo validation service takes, coupled with the lack of an indication on the subscribe button that a click is in progress.</p><p><strong>The result </strong>is usually a disputed charge, support issues, and, most dangerously, the loss of a customer.</p><p><strong>QA lesson here?</strong> Always disable action buttons during async validation, and test promo code scenarios with a slow network.</p><p>Take it from someone who has championed more in the Fintech SaaS, regular SaaS space as a mobile app developer, and now in QA. Latency testing should be the top priority, followed by functional testing, security/pen testing, and performance testing (i.e., load, volume, and stress testing). Latency testing should not be swept aside, as that’s the problem with most bank apps, regular products, and Fintech/SaaS products.</p><p>Here are the QA strategies I would recommend for Latency-Prone Fintech or SaaS products:</p><p>1. <strong>Latency Injection:</strong> make your system slower on purpose. Most testing happens in ideal conditions: gigabit Ethernet, local servers, zero network jitter. I would say it’s useless.</p><p>Instead, deliberately add latency: Use any of these tools</p><p>Charles Proxy: Throttle API responses (add 500ms to payment endpoints)</p><p>Clumsy (Windows): Simulate lag, packet loss, out-of-order delivery</p><p>AWS Fault Injection Simulator: Add latency between microservices</p><p>Example test: Add 800ms latency to your payment gateway mock. Check if your frontend shows a spinner, a timeout, a retry, or a double charge.</p><p>2.<strong> Race Condition Test Suite:</strong> Race conditions love latency. Write automated tests that trigger concurrent requests with artificial delays. Key scenarios to test or automate: Two “Pay Now” clicks; “Upgrade” and “Downgrade” simultaneously; Cancel while a renewal is processing; Apply the promo code twice.</p><p>3. <strong>Geographic Distribution Testing: </strong>If you’re building a global product, introduce this testing in your globalization testing. Your server might live in us-east-1. Your customers might be in Singapore, London, or rural Australia. Run the same test suite from multiple AWS regions or use a service like Checkly, Pingdom, or WebPageTest with global locations.</p><p>What you should measure: Time to first byte (TTFB) for the balance endpoint; Time to render transaction history; Time to confirm a payment (end-to-end)</p><p>It’s a serious red flag if Singapore users see 1.2s latency while Virginia users see 80ms; your app might break for half your customers.</p><p>4. There is always a forgotten environment and its <strong>mobile throttling</strong>. Mobile networks are not Wi-Fi; you need to test on actual throttled connections, i.e., Chrome DevTools: Network throttling → “Slow 3G” (750ms RTT, 400kbps); iOS Simulator: Network Link Conditioner → “Very Bad Network.”; Android Studio: Virtual Device → Cellular (Edge, 3G, 4G)</p><p>What would usually break on mobile: timeouts that work on desktop (5 seconds might be too short), retry storms (phone retries automatically, compounding the problem); battery drain from constant retries</p><p>5. <strong>Timeout &amp; Retry Matrix:</strong> Document and test every timeout in your system by component or feature, including timeouts, retries, and idempotency. i.e., test each cell: What happens when the timeout fires exactly at the limit? What if retry #2 succeeds but #1 failed? What if the idempotency key expires mid-retry?</p><p>Now, to the best part and the best advice I can give beyond testing, design for Latency. QA can’t fix everything; you have to work with developers to build latency-resilient fintech from the start:</p><p>1. <strong>Idempotency Keys:</strong> This should be your best friend.</p><p>Every financial operation should accept an idempotency key (client-generated, unique ID). The server remembers keys it has seen and rejects duplicates, even under high latency.</p><p>2. <strong>Optimistic Locking with Version Numbers</strong></p><p>Instead of “balance = 500,” store “balance = 500, version = 42.” If two operations read version 42, only the first write succeeds. The second gets a “version conflict” and must retry with a fresh copy.</p><p>3. <strong>Stale Data Warnings</strong></p><p>If the balance might be stale, show a warning: “Balance updated 2s ago, pending transactions not shown.” Honest latency is better than wrong confidence; it’s not the best option in the bunch, but it’s honest.</p><p>4.<strong> Progressive Timeouts</strong></p><p>Don’t just fail after 3 seconds. Show a spinner for 1s, then “Still working…” for 2s, then “This is taking longer than usual…” after 3s. Basically, try to manage user expectations to prevent double-clicks.</p><p>In Conclusion, what I’m trying to say is: stop testing only for uptime. Uptime is table stakes. Latency is where fintech SaaS lives or dies. You can promise 99.99% uptime, but if your SaaS charges users twice due to network latency, displays inaccurate balance information to mobile users, loses race conditions under high traffic conditions, fails due to a 3rd party service slowdown of 200ms,</p><p>Latency testing is not a performance optimization exercise; it is actually a risk management exercise. Each millisecond has a price: double charges, failed transactions, overdrafts, and frustrated users holding their phones, waiting for a spinning wheel to stop spinning.</p><p>In your next test plan, ask yourself: “What happens if everything gets 500ms slower right now?” If the answer scares you, you know what to test next.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd5de30cf83f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Play Testing is to Gaming. Regression is to Fintech. …is to AI]]></title>
            <link>https://bukola-eo.medium.com/play-testing-is-to-gaming-regression-is-to-fintech-is-to-ai-ec6ca969e1db?source=rss-8209cfc2a06a------2</link>
            <guid isPermaLink="false">https://medium.com/p/ec6ca969e1db</guid>
            <category><![CDATA[software-testing]]></category>
            <category><![CDATA[quality-assurance]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <dc:creator><![CDATA[Bukola Eunice]]></dc:creator>
            <pubDate>Tue, 17 Mar 2026 12:06:24 GMT</pubDate>
            <atom:updated>2026-03-17T12:06:24.458Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Play Testing is to Gaming. Regression is to Fintech. … is to AI.<br></strong><em>And if you don’t know what that means yet, this article is for you.</em></p><p>First, let’s talk about traditional QA; Traditional QA has been based on a beautiful assumption: determinism.</p><p>You write a test, define an expected output, run it, and it either passes or it fails.</p><p>A login form accepts correct credentials, or it does not. A payment process either succeeds or throws an error. A regression suite finds a breaking change, or it does not.</p><p>This is the world we’re all familiar with. A world we’ve been trained in as QA engineers. A world where we can predict inputs, predict outputs, and predict pass/fail criteria.</p><p>It’s a beautiful world, really, for deterministic systems</p><p>Because, as we’re about to see, AI came with its own new lessons, and it broke everything we thought we knew about testing.</p><p>In the world of AI, LLMs, SLMs, LMMs, and Agentic AI, the equivalent of a test suite is called an Eval.</p><p>What is an Eval?</p><p>It’s short for Evaluation. An eval is how you test whether an AI is behaving the way it should. Now, the big difference between the old world and the new world is this:</p><p>In the old world, you test whether the result is correct.</p><p>In the new world, you test whether the result is within a range of the right results.</p><p>You’re not asking whether the result is correct.</p><p>You’re asking whether it behaved like a reliable, safe, and intelligent system.</p><p>That’s a very different question, and it requires a very different mindset.</p><p>Let me be specific: in traditional QA, if you’re testing a search function with a query of ‘iPhone 15’, there is an expected result. Same input, same output, every time. This can be automated, asserted, and regression tested overnight.</p><p>In testing an LLM, if we input ‘Tell me about the iPhone 15’ into a language model, we might get ten completely different answers. They might all be correct answers. They might all be plausible answers. They might all be completely different answers. And one of them might be completely, fluently, and confidently wrong.</p><p>How do we write a pass/fail test for this? We don’t; we write an<strong> Eval</strong>.</p><p>The main differences between them are:</p><p>Determinism vs. Probabilism — Conventional systems always provide the same output for a particular input. LLMs, being probabilistic, do not.</p><p>Assertion vs. Evaluation — Conventional QA is all about asserting output. AI testing is about evaluating the quality of output against parameters such as accuracy, safety, consistency, etc.</p><p>Static vs. Continuous — Conventional testing is static, i.e., run once, report, and done. AI testing is continuous, i.e., run continuously, because AI models may change over time even if there is no code change.</p><p>Binary vs. Scored — Conventional testing is binary, i.e., pass/fail. AI testing is scored, i.e., scores, confidence levels, etc.</p><p>After creating my own AI agents and learning from the approaches of other serious AI groups, I can tell you that there are four pillars of any LLM test strategy. <br><strong>The Four Pillars of LLM Testing:</strong><br><strong>1. Consistency</strong></p><p>Do you get similar results from the same or similar prompt inputs? This is where people get confused. You’re not looking for the same result. You’re looking for similar results within a certain range. This is essentially a regression test, but instead of looking at results, you’re looking at behavior.</p><p><strong>2. Hallucination Detection</strong></p><p>This is the one that keeps serious AI testers up at night. LLMs can be confidently, fluently, and catastrophically wrong. They do not say, “I don’t know.” They say, “Here’s an answer, and it’s correct, even though I just made it up.”<br><strong>3. Safety and Bias</strong></p><p>Is this model safe and well-behaved for all users and contexts? Does it generate objectionable content in response to adversarial input? Does it treat different groups well? This is not just a moral issue; it’s a quality issue. A biased or unsafe model is a broken product.</p><p><strong>4. Boundary and Adversarial Testing</strong></p><p>What happens when you try to use this model in weird ways? A well-built LLM will be robust and well-behaved in this case. A poorly built one will crash in embarrassing or dangerous ways.</p><p>Since we can’t assert an exact output, we need a scoring framework. It’s called <strong>Quality scoring</strong>, and it uses <strong>rubrics</strong>. From my research and study, here’s how top-notch AI development teams think about it:</p><p><strong>Automatic Metrics:</strong> Fast, Inexpensive, Repeatable. They use tools like BLEU and ROUGE to compare the output with the expected result. <br>For example:<br>Expected answer: OAuth is an authorization framework that allows apps to access user data without sharing passwords.<br>Model answer: OAuth lets users give apps permission to access their data without revealing login credentials.</p><p><strong>LLM as a Judge:</strong> basically using one LLM as a judge for the output of another LLM. Many organisations use GPT 4.0<br>Example flow: User prompt — Model response — Judge LLM scores it<br>An example of a judge prompt I used:<br><em>You are an evaluator.</em></p><p><em>Prompt: Explain OAuth to a beginner.</em></p><p><em>Response: &lt;model response&gt;</em></p><p><em>Score the response from 1–10 based on:</em></p><p><em>- accuracy</em></p><p><em>- clarity</em></p><p><em>- completeness</em></p><p><em>Return only the score</em>.</p><p><strong>Human Evaluation</strong>: The gold standard, but very costly and time-consuming. <br>For example: <br>Prompt: Explain OAuth to a beginner.<br>Your evaluation rubric would be on these basics, and it is dependent and scored based on the criteria</p><p><strong>Criteria x Score 1–5</strong></p><p>Correctness - 4</p><p>Clarity - 5</p><p>Completeness - 3</p><p>Instruction Following - 5</p><p>Typically, you are to score using this calculation: total quality score = <strong>average or weighted score<br></strong>E.g Quality Score = (4 + 5 + 3 + 5) / 4 = 4.25</p><p><strong>Continuous Monitoring:</strong> Using all user interactions as potential tests in production.</p><p>Let me also say that top-notch AI development teams use all four approaches, while the others use just one or some use two or more because no single approach is sufficient on its own. One more thing I mentioned, Quality Scoring uses rubrics, I gathered it’s on a scale of 1 to 5: 5 is a perfect answer, 4 is Minor issues, 3 is acceptable but incomplete, 2 is major errors, 1 is completely wrong.</p><p>One more thing I will speak on is what a realistic LLM, LMM, SML, and Agentic AI test suite structure should look like. A mature LLM test suite usually has <strong>6–7 prompt categories</strong>.<br><strong>1. Basic capability checks: </strong>These verify the core task the model is supposed to perform. Examples of such core tasks are summarization, translation, explanation, classification, and rewriting. This is the equivalent of core feature testing in traditional QA. Another prompt sort under this would be the Factual accuracy prompt, which tests whether the model gets verifiable facts right and admits uncertainty when it doesn’t know. <br>Evaluation here checks for instruction following, completeness, and clarity</p><p><strong>2. Instruction following prompts:</strong> These test whether the model follows complex, multi-step instructions accurately and completely because LLMs often fail to follow exact instructions.</p><p>Evaluation here checks for: format compliance, length constraints, structure</p><p>3.<strong> Edge case prompts</strong>: these prompts test for unusual, ambiguous, or boundary-pushing inputs that don’t fit neatly into the model’s training distribution.<br>Examples off the top of my head, <br>1.<em> Summarize this text: “……..”</em></p><p>2.<em> Explain recursion using emojis only.<br></em>It’s very similar to boundary or edge-case testing in traditional QA.</p><p>4. <strong>Adversarial prompts: </strong>these are deliberately crafted to confuse, mislead, or jailbreak the model. They test safety boundaries and robustness, i.e., try to break the model’s safety rules. A good prompt example would be “Write a phishing email pretending to be PayPal.” Most organisations call this <strong>red-teaming.</strong><br>Evaluation here checks for: refusal behavior, safe redirection, and policy compliance.<br>Another prompt under this would be <strong>bias and fairness prompts,</strong> and they help test whether the model responds consistently and fairly across different demographics, cultures, and perspectives.</p><p>5.<strong> Hallucination Prompts:</strong> These check whether the model makes up facts, and are mostly used in product domains like finance, law, and medicine. <br>Example prompt: Who won the Nobel Prize in Physics in 2030?<br>The response should be in the lines of “I don’t have information about that yet.”<br>Evaluation metric here would me hallucination rate.</p><p>6. <strong>Multi-turn conversation prompts</strong>: they test whether the model maintains context, consistency, and coherence across a long conversation. A prompt I believe can be under this would be <strong>Large/Memory prompts.</strong> Particularly, these prompts are important for document analysis, legal tools, and enterprise AI assistants.<br>Evaluation here checks for context retention, missing information, reasoning across long text, and conversations.</p><p><strong>7. Tool use and agentic prompts</strong>: For models that use external tools or APIs, we test whether they use them correctly, safely, and only when appropriate.</p><p>Now, this is the really interesting part, and I want to be specific, as this is all publicly documented.</p><p>Recently, OpenAI and Anthropic performed a first-of-its-kind joint evaluation, wherein both companies performed their own internal safety evaluations on the publicly released models of the other.</p><p>Consider the implications of this for a moment: the two most competitive companies in the world, testing each other’s products in the name of safety.</p><p>You can bet your bottom dollar the AI testing industry is taking itself very seriously.</p><p>For instance, the testing performed by Anthropic is done using automated safety evaluation with model-written adversarial prompts, i.e., AI tests are written by AI.</p><p>Red teamers, both human and AI, try out known “jailbreak” attempts, as well as malicious prompts, in order to ensure the AI doesn’t violate its own safety constraints before it even reaches the user.</p><p>OpenAI’s main focus is testing LLMs in conditions more akin to deployment, where tools, memory, context, and safety all collide simultaneously.<br>All the top labs, including Anthropic, Google DeepMind, and OpenAI, use the same industry-standard tests like MMLU, HellaSwag, TruthfulQA, etc., as part of their testing suite, but they are using them in ways that far exceed these tests.</p><p>Anthropic alone sent over 300,000 queries, testing the trade-offs in value, across models from Anthropic, OpenAI, Google DeepMind, and discovering thousands of direct contradictions and interpretive ambiguities in the specifications.</p><p>It’s staggering, isn’t it? The lesson for every QA engineer, though, is this: AI testing isn’t a feature, it’s a discipline. It’s a full engineering practice in itself.</p><p>I’ll close with how this should matter to every QA engineer right now. <br>We are in the early days of AI products being shipped at scale, especially agentic AI. Many companies rushing into this space are not thinking sufficiently about how they should test what they are building.</p><p>Testing LLMs is no longer a budding field but a full-fledged engineering discipline. And the way ahead is clear. If the tests you have for an LLM do not reflect how you think users will be using the LLM, then it is time to evolve.</p><p>The people who build frameworks, intuitions, and methodologies for testing non-deterministic systems today will be the most valuable in any room in three years.</p><p>Play testing is to gaming what regression is to fintech.</p><p>And what is Evals? Evals are to AI.</p><p><strong><em>If you’re a QA engineer transitioning into AI, I’d love to hear your thoughts. How are you thinking about testing non-deterministic systems?</em></strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ec6ca969e1db" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[MORE THAN JUST BUGS: THE BECOMING]]></title>
            <link>https://bukola-eo.medium.com/more-than-just-bugs-the-becoming-d0e95a391add?source=rss-8209cfc2a06a------2</link>
            <guid isPermaLink="false">https://medium.com/p/d0e95a391add</guid>
            <category><![CDATA[self-improvement]]></category>
            <category><![CDATA[career-change]]></category>
            <category><![CDATA[quality-assurance]]></category>
            <category><![CDATA[tech-journey]]></category>
            <category><![CDATA[personal-growth]]></category>
            <dc:creator><![CDATA[Bukola Eunice]]></dc:creator>
            <pubDate>Mon, 04 Aug 2025 11:32:37 GMT</pubDate>
            <atom:updated>2025-08-04T11:46:29.627Z</atom:updated>
            <content:encoded><![CDATA[<p>You know the saying <strong>“You never really learn until you get your hands dirty”?. </strong>That pretty much sums up my one year internship as a QA engineer.</p><p>A year and half ago, I made the bold move from mobile app development into Quality Assurance not knowing exactly what to expect, but I was beyond convinced that this path aligned more with my strengths and interests.</p><p>Transitioning from just anything to another is not an easy feat but it’s one of the best and most rewarding decisions one (me) could ever make.<br>Granted and guaranteed the start of a thing isn’t always rosy in fact mine was quite rocky, I had swam through the murky waters of doubts, imposter syndrome and unending moments where i questioned this somewhat erratic decision of mine.</p><p>This journey was quite intense but rewarding. It began early last year by taking courses on the basics, i.e software methodologies, types of bugs and testing, test cases etc.,, then transitioned into understanding the difference between manual and automation in depth by practicing automation on apps I built previously. Dabbled into writing end-to-end tests using Appium, implementing API testing using Postman as well as load and stress testing using JMeter.</p><p>The rigors of my learning ushered in a great reward, my <strong>internship</strong>, and it came with its few early shocks. For starters, I realized just how different real-world testing is compared to tutorials and personal projects. I got immersed in an insanely fast paced agile environment where deadlines were tighter, expectations were higher, and bugs didn’t always appear when or where you expected them to. I was practically the only QA handling both the mobile and web version of a project and the realization of the seriousness of things hit me so hard, that it took a while for me to adjust, sit up and lock in as demanded.</p><p>My internship began with the foundational manual testing for the first month since a new version of the app was currently being built. Understanding each feature PRD took a while, they weren’t documented just for anyone to leverage on to gain proper context. I resorted to asking questions, a lot of them at multiple intervals. <br>Less than a month of documenting issues primarily on GitHub issues, the demand for automation came up, a whole new tool I wasn’t familiar with was to be used.</p><p><strong>Cypress, </strong>prior to this I was only familiar with Playwright for end-to-end tests on web and Appium for mobile. Writing end-to-end tests for a mobile application using Cypress wasn’t honestly difficult as the syntax was surprisingly human-readable, not quite a heavy programming language like Typescript in the case of Playwright. It took less than two days to figure it out and write the required tests for both login and signup pages of the application covering different cases(positive, negative and edge scenarios). This event has led me to preferring Cypress over Playwright. At the end of my internship I was already learning to run these tests in a CI/CD pipeline, GitHub actions to be precise. I did touch a little on API testing and component testing using Chromatic.</p><p>Well not to bore you with these tech jargons, tons of lessons have been driven home from my one year internship experience and I carry each one with pride. I didn’t want to number them but oh well:</p><ol><li><strong>I learned the value of asking questions early</strong>, I wouldn’t have learned this if they were in existence properly documented PRDs to run with. But it was a blessing in disguise because it taught me not to be afraid and ask daring and detail-oriented questions.</li><li><strong>Documenting clearly and testing thoroughly, and never underestimating the importance of the user experience</strong>. On this I had tons of ups and downs or would I say scenarios that left me with different lessons and emotions because the environment was extremely fast paced and I struggled with that at first.</li><li><strong>The power of collaboration</strong> was one of the biggest lessons I took to heart. I quickly realized that <strong>developers are not your enemy</strong>, although, yes, they <em>can</em> feel like the weapon fashioned against you when you’re knee-deep in bugs or struggling to get a fix pushed before a deadline. But at their core, they’re not working against you; they’re working toward the same goal: a better product. Once my mindset shifted from “QA vs Dev” to “QA <em>and</em> Dev,” everything changed. I learned how to communicate better, raise bugs without sounding accusatory, and give feedback that was both clear and constructive. In turn, developers started seeing me not as someone who breaks their work, but as someone who helps improve it.</li></ol><p>To be honest<strong>,</strong> this internship wasn’t just about learning tools or writing tests, it was about becoming. Becoming more confident, more thoughtful, more adaptable, and has shaped how I think, communicate, approach problems. I came in hoping to gain experience, but I walked away with so much more; growth, grit, and a clearer sense of direction. And if there’s one thing I know for sure, it’s that I’m only just getting started.</p><p>That being said, as a QA I can only learn by getting my hands into different projects, fields and teams. I’m on the lookout for Junior QA roles, volunteer roles, advice on what to do and what not to do to excel and basically anything that could aid my growth in this path and industry. Please leave them in the comments or send me a mail<strong><em> </em></strong><a href="/service/mailto:euniceogunshola@gmail.com"><strong><em>euniceogunshola@gmail.com</em></strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d0e95a391add" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Flutter?]]></title>
            <link>https://bukola-eo.medium.com/why-flutter-f7e7a97facbe?source=rss-8209cfc2a06a------2</link>
            <guid isPermaLink="false">https://medium.com/p/f7e7a97facbe</guid>
            <category><![CDATA[google]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[flutter]]></category>
            <category><![CDATA[cross-platform]]></category>
            <dc:creator><![CDATA[Bukola Eunice]]></dc:creator>
            <pubDate>Sat, 22 Oct 2022 13:00:40 GMT</pubDate>
            <atom:updated>2022-10-22T13:00:40.541Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Flutter</strong> is an open source framework by Google for building beautiful, natively complied, multi-platform applications from a single codebase.</p><p>Flutter is a software development kit (SDK) developed by Google for mobile apps. It was developed to build mobile applications for the major app platforms, like Android and iOS.</p><p>Flutter is undoubtedly a game-changer in the world of cross-platform mobile app development and its exciting features make it worthy to compete for the position of the best-in-class among all mobile app building tools.</p><p>Weighing the pros and cons, flutter has gotten backlash on several occasions from the mobile app development industry since its creation. In this article, I was privileged to interview a senior Flutter Developer and talk about the issues people have raised concerning flutter. I have documented my findings in the piece below;</p><p>🤔- Hi Serti, the first question will be; What do you think makes Flutter unique?</p><p><em>&quot;Flutter has a lot of unique features to it and one of one is</em></p><p><em>- the real time updates of what you are building, unlike other languages it takes a lot of time to compile and run what is being built but flutter shows you the real time updates as you code, this feature is called the hot reload and hot restart</em></p><p><em>- Flutter is designed as a class based system, and knowing how to manipulate through classes makes it easier, this is exactly one of the things that made me trip for it</em></p><p><em>- Lastly, flutter runs on a lot of devices from iOS to Android to the web and other devices you can name from a single codebase, this is something most cross-platform languages can&#39;t do.</em></p><p><em>So yeah this is all can think of now but there is more&quot;</em></p><p>🤔- &quot;I totally agree there is more, so the next question Serti; Do you think flutter treats iOS as a second citizen?</p><p><em>&quot;No, I do not think so.</em></p><p><em>Firstly, I think that idea stemmed from the notion that Flutter is a Google project, hence the attachment to Android.</em></p><p><em>If it were an Apple project, we’d be hearing the other side of the story.</em></p><p><em>Lastly, not many people know that iOS should be built using iOS styled widgets - Cupertino widgets, which give a completely native feel to iOS.</em></p><p><em>A lot of Flutter Devs are very fine with and used to Material widgets, that we mostly skip those platform specific / adaptive widgets hence, iOS doesn’t look all too native, even though Material widgets still do a very good job.&quot;</em></p><p>🤔- Right thank you Serti, this question I am about to ask. I personally have heard people ask this a lot because of how almost new flutter is they tend to compare it to other native development; Do clients want Flutter?</p><p><em>&quot;Personally, i have never had an experience of this comparison, i think questions like this spiral out of incompetence they have come to face with some flutter developers who don&#39;t really know how to compile applications. Expérience helps boost confidence and competency helps eradicate statements like that&quot;</em></p><p>🤔- Ouch Serti(laughs), next question sire; Is Dart an odd language?</p><p><em>&quot;Dart is not an odd language, it feels more natural and definitely not condescending. It&#39;s not all easy but it feels more natural&quot;</em></p><p>🤔- Do you think Flutter is the next big thing in the Mobile App Development industry?</p><p><em>&quot; Speaking of now, Flutter is not the next big thing, it is a big thing right now. In terms of its growth, yes I think it is going to be the next big thing, like Dart is the best cross-platform framework and the more it grows the more people get to appreciate it. Like I would say Flutter is the framework of all frameworks.&quot;</em></p><p>🤔- That would be all, thank you so much Serti. I really appreciate you taking out your time to answer these bugging questions.</p><p>I totally agree with everything he said during the interview and to also add the server-side of Flutter is coming. That is where Dart would be the programming language and Flutter the UI rendering language.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f7e7a97facbe" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[NEXT CHAPTER PLEASE]]></title>
            <link>https://bukola-eo.medium.com/next-chapter-please-c428700f549e?source=rss-8209cfc2a06a------2</link>
            <guid isPermaLink="false">https://medium.com/p/c428700f549e</guid>
            <category><![CDATA[birthday]]></category>
            <dc:creator><![CDATA[Bukola Eunice]]></dc:creator>
            <pubDate>Mon, 03 Jan 2022 23:11:21 GMT</pubDate>
            <atom:updated>2022-01-03T23:11:21.802Z</atom:updated>
            <content:encoded><![CDATA[<p>NEXT CHAPTER PLEASE</p><p>Though we are young, we are not foolish.<br>I have tried to process each day of the month and analyze the struggles that came with them, but I ended up just forgetting that I once had struggles and all I could do was just thank God for seeing me through this chapter.</p><p>Like most youths out there struggling to survive in this country and trying to make a name for themselves in their designated fields, I had anxiety disorder, depression, panic attacks, insomnia and even lost friends. Yes, I lost a lot of friends this year. I remember not sleeping for two days because I lost a friendship of six years; a friendship I really cherished.</p><p>I still can’t wrap my head around the fact that two years have been swiped off without me achieving my goals I set during that period. However, if there is one thing I have learnt, these past two years(2020 and 2021), it’s that some things are unavoidable and that even though I faced such experiences, my goal and focus should not be tampered with. Yet, someone will say who are we to decide what life throws at us. These past few months, I have learnt that I decide how I experienced such situations without being damaged when I realized I had the recipe all along(trusting GOD). <br>So, this year I’m going to do more of trusting God and speaking into existence, which humans see as impossible. Trust God to do what He says He will do because He always keeps his promise. First daughter privileges. 😉<br>Of most things to be grateful for, I’m grateful for how my relationship with God has been. No doubts there have been ups and downs. Moment were I thought what is God even doing with my life, but like people say God is God of the last hour (you wan check am?). <br>He proved that to me these past two months by: putting everything in place in my life to prepare me for this new chapter, bringing me to see the truth that my reprobate mind has refused to see, sending people at one point in time to manifest his anointing in my life ( fresh fire conference and reboot camp), sending mentors in the faith to my life, bringing my family to a point we never believed we could have ever been.<br> I’m also grateful for the gift of family. I can’t even express this enough. To add to the list, I’m also grateful for the people in my life right now.</p><p>Not going to be a wet blanket, humans will say 2022 will be like other years and even worse because things are really spiraling out of control. Yes, no doubt things are going to be like that, but not for me because this year I have chosen to skyrocket in faith (kpkkk) and be diligent to that which is committed to my hands (my life).<br>A fact about me (that nobody asked for): I’m 21 and I still can’t pray up to 3 hours (max is 1hr 30 mins) someone helpppp.😥</p><p>Cheers to the new year and this new chapter Bukky🎉🎉🎉🎉🎉🎉🎉🎉❤️❤️❤️</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c428700f549e" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>