<?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 Dilip Thomas Ittyera on Medium]]></title>
        <description><![CDATA[Stories by Dilip Thomas Ittyera on Medium]]></description>
        <link>https://medium.com/@dilipti?source=rss-3d3b1fa75874------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*5B4lbS3JnOcb-OrH.jpeg</url>
            <title>Stories by Dilip Thomas Ittyera on Medium</title>
            <link>https://medium.com/@dilipti?source=rss-3d3b1fa75874------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 14:01:41 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@dilipti/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[Why are we building CogniSwitch.ai]]></title>
            <link>https://dilipti.medium.com/why-are-we-building-cogniswitch-ai-78f4062da6a1?source=rss-3d3b1fa75874------2</link>
            <guid isPermaLink="false">https://medium.com/p/78f4062da6a1</guid>
            <category><![CDATA[rags]]></category>
            <category><![CDATA[knowledge-graph]]></category>
            <category><![CDATA[genai]]></category>
            <dc:creator><![CDATA[Dilip Thomas Ittyera]]></dc:creator>
            <pubDate>Wed, 10 Apr 2024 06:24:46 GMT</pubDate>
            <atom:updated>2024-04-10T06:24:46.164Z</atom:updated>
            <content:encoded><![CDATA[<p>In recent times, not a day passes by for the Generative AI ecosystem without the breaking story of a new model or a paper outlining a new approach or startups surfacing from stealth mode with multi-digit rounds of funding. Amidst all this uplifting stories, the results of surveys among Fortune x companies by some of the leading analysts show that less than 20% of all proof of concepts and pilots in this space has gone into production. This is critical for these enterprises as well as for the startups and the VCs who have taken a bet on these startups.</p><p>While a ton of new technologies and approaches have surfaced, the mob attitude that it has to be a particular technology or a specific approach that will work is killing the collaboration and collective innovation required to move forward some of the very exciting possibilities for progress that mankind has encountered. A case in point is using embeddings and vector databases and a ton of variations in retrieval augmented generation (RAG) in conjunction with LLMs and hoping that hallucinations and serious reliability issues will somehow get eliminated.</p><p>When we started working on the very first versions of CogniSwitch much before GenAI made it’s appearance, we had set ourselves the vision of helping organizations create digital twins of their knowledge. There is a key reason for us to start thinking that knowledge would now suddenly become important for organizations such that they will worry about creating digital twins of it — the reason was the fact that many teams were becoming hybrid. By hybrid we mean teams being made up of human being and artificial beings (AI).</p><p>Many teams are yet to reconcile to this hybrid nature of their teams. And even if they did, it would be impossible to utilize the current proven process of on-boarding members to a team given that the process was built with humans in mind. We are convinced that organizations need to adopt newer means of doing this with AI members of your teams. And when this is not done, sh*# happens as in the case of the now (in)famous of what happened with a customer of Air Canada!</p><p>Every credible organization has their product, service and process knowledge as well as policies, regulations and other SOPs documented. And this documentation is done primarily using natural language. This is the foundation on which any organization operates. And it is critical that we bring this to bear on any member in our teams, especially the hybrid ones.</p><p>We built CogniSwitch (CS) to consume these important natural language content and gather the facts and knowledge contained within and organize in a structured knowledge base. We also built CS to be able to provide reliable information when team members seek these facts and knowledge relevant to the tasks/actions they are performing, especially the AI members in the team. CS provides both the ingestion and retrieval mechanisms using APIs so that this can happen in a seamless fashion with existing and new applications that the organization utilizes.</p><p>The ingestion of your natural language sources of data, transforming it to a enabling structure and persisting it in a knowledgebase utilizes an automated, optimized and configurable pipeline. This speeds up the process of getting the critical data into a shape and form directly accessible by AI and other applications. The fact that the data comes from facts and knowledge shared by human experts and curated by them and that the data retrieval process is based on a deterministic algorithm ensures that the retrieved data is reliable and relevant for the task at hand. Once this is set up, it becomes much easier to retrieve the relevant knowledge require to respond to a query or feed a creative Generative AI pipeline to produce a desired output.</p><p>For those of you with a technical bend of mind, CogniSwitch internally utilizes a set of capable mechanisms like embedding, LLMs, NLP, ontology/taxonomy and knowledge graph logic and rules among other things in the ingestion pipeline. CS also uses natural language, knowledge graph traversal, flow and decision engines as part of the platform to ensure reliability.</p><p>In the case of GenAI, the most important point to note is that while CS utilizes most of the current approaches including an external data source, the unique approach and combination of these technologies is what makes it work. In this case, natural language data sources and LLMs dominate the system, but utilizing the right component for the right task in the system is crucial. We chose to utilize the highly capable LLMs for it’s strengths in the ingestion pipeline and avoided using them in the retrieval pipeline where their weaknesses could destabilize and make most systems unreliable.</p><p>In case you would like to understand more about our approach or try it out on any of your use cases, do reach out to us at dilipti@cogniswitch or register at <a href="https://cogniswitch.ai/">https://CogniSwitch.ai</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=78f4062da6a1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Large Language Models and Knowledge Graphs: A Journey Towards Collaborative Intelligence]]></title>
            <link>https://dilipti.medium.com/large-language-models-and-knowledge-graphs-a-journey-towards-collaborative-intelligence-da128e297de6?source=rss-3d3b1fa75874------2</link>
            <guid isPermaLink="false">https://medium.com/p/da128e297de6</guid>
            <category><![CDATA[large-language-models]]></category>
            <category><![CDATA[knowledge-graph]]></category>
            <category><![CDATA[vector-database]]></category>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[embedding]]></category>
            <dc:creator><![CDATA[Dilip Thomas Ittyera]]></dc:creator>
            <pubDate>Wed, 28 Jun 2023 20:39:40 GMT</pubDate>
            <atom:updated>2023-06-28T20:39:40.458Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*3LzPRESriTVt9LuV.png" /></figure><p>There’s been a lot written and tweeted about Generative AI (GenAI) powered by transformer based large language models (LLM). OpenAI in sync with Microsoft clearly leads the table with Google and other big players and a bunch of innovative open source options in the fray.</p><p>As with any other technology with potential for disruption in varied fields and diverse candidate use cases, attendant hype is in step, causing a lot of confusion.</p><p>In the late 1980s, as a young enthusiast working with Expert Systems, I was captivated by the promise of rules-based symbolic AI. I witnessed firsthand the power of these systems and the transformative potential they held. Yet, a challenge remained — it was impossible to manually capture all the nuanced knowledge of human experts into these systems. Despite all the hype then, there came an AI winter, forcing many of us to pivot.</p><p>Over the next three decades, as I moved up in my career spanning large tech organizations and startups across various geographies, this challenge remained. Efficiently capturing expert knowledge and it’s efficient reuse still begged for a solution. It was not until the advent of transformer architecture and the emergence of models like GPT, combined with the wider use of Knowledge Graphs (KG), that the possibility of an answer to this challenge began to emerge.</p><p>If we leverage their inherent strengths, LLMs can be utilized for analyzing vast volumes of unstructured data, extracting facts and knowledge into a requisite structure that can be organized in KGs. KGs enrich this information by drawing connections, filling in gaps, and supplying contextual understanding. In return, KGs offer LLMs a reliable, validated source of information, grounding the outputs of the LLM on facts and verified knowledge. This helps LLMs shed hallucinations and reduce errors. KG based embeddings can help refine LLMs with specific domain and enterprise knowledge. This mutual interdependence sets up a virtuous cycle of continuous learning and improvement. The interplay between LLMs and KGs creates a dynamic system that evolves and improves over time.</p><p>However, in this context, it is essential to address a common misconception. Many of us assume that creating embeddings of chunked up documents and storing them in a vector database (vDB) can solve the hallucination problem in LLMs. While it seems logical, let us scrutinize this assumption. Even the best in class LLMs, despite their high-dimensional vector spaces, still produce hallucinations. These are highly capable models compared to the lesser models we often utilize in creating embeddings for vDBs. Despite their advantages, hallucinations persist for the SOTA LLMs. So, when we retrieve chunks of content from these vDBs that operate on probabilistic algorithms and feed it to an LLM, expecting hallucination-free outputs might be overly optimistic.</p><p>This does not devalue the use of vector databases and embeddings — in fact there are synergies between vDBs and KGs that can be leveraged. Intention is to highlight that eliminating hallucinations is a complex task, needing more than just embeddings alone. It also reinforces the importance of KGs as a validated knowledge source to anchor LLM outputs and ensure their accuracy.</p><p>The vision of integrating LLMs and KGs goes beyond managing knowledge within an organization. It is about enabling AI as a collaborator in human progress, building on the collective spirit of human knowledge sharing. Remembering that successful collaborations involve thoughtful design, robust security and privacy focus, and ongoing management is also essential.</p><p>Reflecting on my journey from Expert Systems to LLMs and KGs, it is clear we have made significant strides that benefit humanity. It also underscores our ability to learn, adapt, and innovate, and fuels excitement for the next phase of our AI journey.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=da128e297de6" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>