I'm Anuj Mendhule, Azure AI Solutions Architect. For 9+ years I've taken
AI and data platforms across the gap where most projects die:
between the demo and production. RAG pipelines that stay
accurate, integration that doesn't page anyone at 2 AM, and security
done right from day one.
Conversational AI over your real data, designed for accuracy,
traceability, and cost. Not just a demo.
RAG architecture on Azure AI Search (vector + semantic)
Agentic workflows with LangGraph / LangChain
Azure OpenAI with Managed Identity & Key Vault
Ingestion for PDF, DOCX, PPTX, XLSX at scale
SVC-02
Data platforms & integration
Multi-source data platforms with the operational discipline of
software: CI/CD, environments, reconciliation.
Azure Data Factory at scale (80+ pipelines)
Event-driven ingestion: SFTP, REST, S3, webhooks
ARM + Azure Pipelines across dev/test/UAT/prod
Warehouse design, ETL/ELT, data quality frameworks
SVC-03
Product engineering
When the problem needs a product, not a diagram, I design, build,
and ship it end to end. See the two below.
React / TypeScript + FastAPI or .NET backends
Firebase: Auth, Firestore, Hosting, App Hosting
React Native / Expo + installable PWAs
From first wireframe to production deploy
selected enterprise work
Systems running in production today.
CASE-01enterprise ai
Document Intelligence & RAG platform
Azure OpenAI · AI Search · LangGraph · Functions · Cosmos DB · FastAPI · React
End-to-end RAG pipeline: vector + semantic retrieval with GPT-4 and LangGraph orchestration for conversational Q&A over enterprise document repositories.
18+ Azure Functions ingesting PDF/DOCX/PPTX/XLSX: chunking, embeddings, vector indexing, and a Cosmos DB change-feed feedback loop.
Router-based query classification in a React + FastAPI platform, plus a 32-component shared UI library published via Azure Artifacts.
CASE-02data platform
Commodities data aggregation platform
Azure Data Factory · Functions (.NET 8) · Logic Apps · SQL Server · ARM · Azure Pipelines
84 ADF pipelines, 60+ triggers: ETL from SFTP, CME, USDA, HubSpot, AWS S3 and Dropbox into one governed platform.
52-endpoint REST API with multi-tenant row-level security, token auth, and Key Vault-managed secrets.
Financial-grade integrity: daily FCM transaction reconciliation across 3 brokers; full dev/test/UAT/prod CI/CD via ARM templates.
CASE-03ai + analytics
ISP analytics & intelligent data platform
Python · FastAPI · ChromaDB · Ollama · React · SSIS · Power BI · PostgreSQL
RAG chatbot over 3,900+ SQL objects, Power BI reports and SSIS packages, with hybrid vector + keyword search and SSE streaming answers.
Cross-domain lineage graph: an indexer parsing SQL DDL, SSIS XML and .pbix, visualized as an interactive React Flow dependency map.
11-stage incremental ETL, 37 materialized views, and a 61-view data-quality exception framework.
products i've built & run
Side projects, shipped like products.
Both are live, self-funded, and built end to end: architecture, design,
code, and deployment. They're the clearest proof of how I work.
live · consumer app · pwa + android
Homebrew
homebrew.imanuj.com
A pour-over coffee companion that treats brewing like an engineered
process. A deterministic recipe engine turns method,
roast, process, and freshness into exact temperature, ratio, and pour
schedule, reproducible every morning with no AI guesswork. A live
brew clock then guides the pour second by second.
Tip · pour in slow spirals, center-out. Keep the bed flat.
live · b2b saas · multi-tenant
DineIn
dinein.imanuj.com
Design-first QR menus for cafés and restaurants. Guests scan a table
QR and get a beautiful, branded mobile menu; owners
manage brand, branches, menus and events from a self-serve admin,
and export print-ready PDFs with their QR embedded.
The complete picture: enterprise Azure AI projects, the data
platforms behind them, and the products above. Leave your details
and it downloads instantly.
▸ goes only to me · no mailing list · i usually reply within a day
contact
Let's build something that ships.
Open to consulting and architecture engagements in AI platforms, RAG
systems, and data integration, and to conversations about full product
builds. The fastest way to reach me is LinkedIn.