I build the knowledge layer that AI systems retrieve from. Enterprise corpora are written for human readers, which is why retrieval over them performs badly — my work is turning a large product documentation corpus into something an LLM can query accurately, and then measuring whether it actually works.
14 years across Actian, Adobe, ABB, and Aristocrat Technologies, from DITA migrations and taxonomy design through to RAG pipelines, retrieval evaluation, and Model Context Protocol (MCP) servers.
Documentation now has two audiences: the people reading it and the systems querying it. I build for both, and I measure the second.
| Project | What it demonstrates |
|---|---|
| docs-mcp | MCP server in TypeScript. Five tool handlers exposing a documentation corpus, GA4 content gaps, and Jenkins CI status to Claude in one conversation. The base retrieval layer |
| knowflow | Evolution of docs-mcp — adds a RAGAS-style evaluation loop (relevance, faithfulness, recall) so retrieval quality is measured, not assumed. The full content-intelligence loop |
| Documentation-AI-Assistant | RAG pipeline and chat UI over a documentation corpus — embeddings, semantic search, and grounded response generation |
| knowledge-graphs-for-ia | Turns documentation into a typed knowledge graph (REQUIRES / COVERS / BELONGS_TO) with graph-traversal retrieval. GraphRAG generation in progress |
| docs-style-guard | Automated writing-standards enforcement — turns a style guide into a check that runs on every change, callable over MCP |
| Knowlayer | Live service and writing on making enterprise content AI-ready, including deep-dives on chunking and knowledge graphs |
| information-architecture-playbook | IA principles, governance checklists, maturity model, and content-modeling templates |
AI-Ready Content Systems — I engineer documentation corpora that LLMs can retrieve and reason over: chunking strategy, semantic metadata, and structured authoring so content performs in a RAG pipeline, not just in a browser.
Retrieval & Evaluation — RAG pipelines with RAGAS-style evaluation measuring answer relevance, faithfulness, and context recall against a repeatable baseline. Evaluation is the part most RAG projects skip, and the reason most of them cannot be trusted in production.
MCP & Tool Design — MCP servers that expose a corpus, analytics, and CI status as tools any MCP client can discover and call.
Information Architecture — The structures that make content findable, reusable, and consistent at scale: taxonomy, metadata schemas, content models, navigation systems, and governance frameworks.
Docs-as-Code & CI/CD — Publishing pipelines (Jenkins, GitHub Actions, MkDocs) that treat documentation like software: version-controlled, validated, automatically deployed.
Analytics-Driven IA — GA4, BigQuery, and content-gap analysis wired into IA decisions: measure what users cannot find, then restructure to fix it.
AI & Retrieval — RAG · Vector embeddings · Semantic search · Model Context Protocol (MCP) · RAGAS evaluation · Knowledge graphs · ChromaDB · pgvector · Prompt engineering · LLM integration
Core IA — Taxonomy & ontology design · Metadata standards · Content modeling · Topic-based authoring · DITA · Single-source publishing · Navigation & findability · Governance frameworks
Tooling & Engineering — Docs-as-code · Python · TypeScript · Jenkins CI/CD · GitHub Actions · MkDocs · Markdoc · AEM · FrameMaker
Developer & API Docs — REST · GraphQL · OpenAPI/Swagger · Postman · Interactive sandboxes · Developer onboarding
Analytics — Google Analytics 4 · BigQuery · Adobe Analytics · Content performance dashboards · SEO
| What I improved | Result |
|---|---|
| Content discoverability | +45% |
| Organic search traffic | +50% |
| Duplicate / redundant content | −35% |
| API self-service success rate | +40% |
| Engineering onboarding time | −70% |
| Documentation build & publish time | −40% |
| Adobe HelpX users supported | 1M+ / quarter |
| Role | Company | Period |
|---|---|---|
| Principal Information Architect | Actian | Apr 2024 – Present |
| Information Architect | Adobe | Sep 2021 – Apr 2024 |
| Senior Technical Writer / Information Architect | ABB | Aug 2018 – Sep 2021 |
| Senior Technical Writer | Aristocrat Technologies | Jun 2012 – Jul 2018 |
🎓 B.Tech, Aeronautical Engineering — R. V. College of Engineering, Bangalore
📜 Prompt Engineering: How to Talk to the AIs (2025) · UX Foundations: Information Architecture · Jenkins Essential Training
Portfolio · LinkedIn · IA Playbook · pandey.bipin2@gmail.com