Published Work
Research
Argues hallucination is epistemological, not engineering. Proposes a Pramana-based AI architecture from Advaita Vedantic and Nyaya epistemology — six irreducible knowledge pathways, each with explicit validity conditions.
For AI Agents
Machine Interface
This site speaks MCP. Agents connect directly — query my portfolio, traverse the knowledge graph, or ask anything. Programmatically.
Streamable HTTP
Building Now
In Public
AI citation verification for practising lawyers. A six-pramana reasoning engine that ships verdicts with evidence, not vibes. Wave 1 shipping to small firms now.
A published skill that teaches AI coding assistants to think in UI/UX principles. 12 principles, 13 reference docs.
This site and an MCP server at mcp.narenkatakam.com — Astro, Cloudflare Workers, Vectorize RAG. Queryable by any AI agent as easily as a human reader.
Case Studies
2018 – 2024
2024
Most recent project (2024). Developer platforms are products too. Built a 9-layer AI-native platform where agents handle code reviews, security scans, and compliance — so engineers ship faster without sacrificing quality.
Insurance
0→1
2022
Data problems are culture problems. Created the Data Trinity framework — availability, accessibility, literacy — to move a manufacturing giant from siloed spreadsheets to data-driven decisions.
Manufacturing
1→n
2021
Pitched a digital twin to compress drug discovery timelines using ML-driven cohort simulation. The vision landed but the project did not get funded. Data governance killed the funding. The ambition was right — the organization wasn't ready.
Pharmaceutical
0→1
2019
What if government worked like an API? Connected siloed agencies through AI agents and conversational interfaces — years before 'agentic' became a buzzword. Then COVID hit, and we pivoted overnight.
Government
0→1, 1→n, n→∞
2019
An airline becoming a super-app had millions of users but zero personalization data. Designed a 3-layer recommendation engine — from basic rules to reinforcement learning — solving the cold-start problem.
Aviation
n→∞
2018
Airlines were losing aircraft to preventable hard landings. Built a data platform combining real-time analytics, predictive maintenance, and an AI chatbot — turning scattered flight data into actionable intelligence.
Aviation
0→1, 1→n, n→∞
2018
Banks were losing to tech giants who understood their customers better. Designed an 'invisible bank' — one that lives inside daily routines, analyzing spending in real-time, anticipating needs before you feel them.
Banking
1→n, n→∞
Build logs & essays
When I ship something, break something, or change my mind — you hear first. Low-frequency. No filler. Unsubscribe any time.
Double opt-in. Unsubscribe any time. No third-party trackers.
