Languages
Python and TypeScript carry most of my work, with JavaScript, Rust, and Java on the side.
Open to full-time roles
Hi, I’m Snehel Basu
AI backend Engineer — LLM systems, multi-agent workflows, and Python backends
I work mostly on the backend of AI products — multi-agent orchestration, retrieval pipelines over knowledge graphs, and FastAPI services built to stay maintainable as they scale. Most recently I built a Graph-RAG knowledge engine and a multi-tenant chatbot platform at Zaito AI, and I care about clean architecture and getting the boring parts right.
Projects
A few things I have designed and shipped lately — products, open source, and one experiment that never quite became a product.
- 01
Multi-Agent Research System
A LangGraph pipeline that runs several agents in parallel to plan, research, and write a report on a topic. Redis and Celery handle the job queue, SSE streams progress live, and E2B sandboxes any code the agents generate.
2026 · AI toolingView on GitHub - 02
AI Trading Arena
Several AI agents paper-trade perpetual futures independently. A pipeline derives technical indicators such as MACD and RSI and feeds them to the model as context before it places a trade, through the Lighter SDK.
2026 · Applied AIView on GitHub - 03
RAG Job Matching Platform
Recruiters post a role and immediately get a ranked list of matching candidates. Built on RAG best practices — query rewriting and metadata filtering — to keep the matches accurate. Won second prize at the CodeConquer ’25 hackathon.
2025 · Hackathon winnerView on GitHub
Experience
Where I have worked, what I owned, and what shipped because of it.
- Nov 2025 — Jun 2026
Graph-RAG & Knowledge Graph Engine · Zaito AI
Backend Developer Intern
- Built an automated collection pipeline that scrapes news articles and web content into an internal AI knowledge graph.
- Designed a multi-agent extraction pipeline that identifies entities and selective relationships in raw text, turning unstructured documents into connected graphs.
- Developed the Graph-RAG natural language search layer, letting users query graph data and get back contextual nodes, relationships, and cited answers.
- Integrated third-party APIs such as HubSpot and engineered the data pipeline to onboard over 10k existing CRM records into the knowledge base.
- Nov 2025 — Jun 2026
Multi-Tenant Chatbot-as-a-Service · Zaito AI
Backend Developer Intern
- Architected the platform from scratch on FastAPI and PostgreSQL, with an agency dashboard for tracking chats, leads generated, and reviews.
- Engineered customisation controls so agencies can tailor chatbot styling, configure custom Q&A datasets, and build dynamic lead-capture forms.
- Implemented domain whitelisting so the widget validates origin domains before serving, preventing unauthorised site-key usage on unapproved sites.
- Nov 2025 — Jun 2026
Title Insurance Assistant · Zaito AI
Backend Developer Intern · Core product
- Contributed to the flagship AI assistant, building document tagging so users can query precisely across selected real estate title documents.
- Built the pipeline that generates a property's initial report in a customisable format with PDF export.
- Executed a large-scale Qdrant migration, re-embedding thousands of production records to move the vector store from OpenAI to Google embeddings.
Skills
The tools I reach for most, grouped by what I actually use them for.
AI & orchestration
LangGraph, LangChain, and PydanticAI for multi-agent systems, tool use, and typed agent flows.
RAG & retrieval
Graph-RAG and vector search across Qdrant and Neo4j, including large-scale re-embedding migrations.
Backend & APIs
FastAPI, Express, and Node.js behind REST and WebSocket APIs, with a bias toward clean architecture.
Async & infrastructure
Redis, Celery, and SSE for job queues and live progress streaming; Docker, Postgres, and Alembic for delivery.
Data & integrations
Web scraping pipelines, third-party API integrations such as HubSpot, and onboarding messy data into knowledge bases.
Interests
Mysteries, mostly. Japanese detective fiction, locked-room puzzles, and one web novel that rewired how I think about worldbuilding.
Lord of the Mysteries 
The Inugami Curse 
Strange Houses 
The Devil's Flute Murders 
And Then There Were None 
The Murder of Roger Ackroyd 
Days At The Morisaki Bookshop 
Murder on the Orient Express
Hover the shelf to stop it. 8 books, mostly detective fiction.
05
Let’s build something together.
basusnehel789@gmail.comOpen to full-time roles — in AI or backend engineering at an early-stage startup, or a project worth building together.