Available for new opportunities

Building thoughtful software for the AI era.

I’m Rohit Raj — a Software Engineer based in Bengaluru, focused on building practical AI systems. I create AI agents, RAG systems, and scalable APIs that solve real-world problems and hold up outside the demo.

See my work Get in touch
Based in
Bengaluru, IN
Focus
AI agents · RAG · APIs
Stack
Python · TypeScript
01 — About

A short version of the story.

Most of my work lives in the unglamorous middle of applied AI — where a model meets a real product and almost everything that can go wrong, does. I build the layer that catches it: agents with tool use, retrieval pipelines that survive bad PDFs, and APIs that don’t fall over when something stops being a demo.

I care about taste as much as throughput. Typography, motion, and information density tell a user as much about a product as the model behind it — so I keep one foot in backend plumbing and the other in interfaces that feel quiet.

When I’m not shipping, I’m reading other people’s source code, benchmarking the latest model the internet pretends will change everything, and quietly resisting the next framework that promises to fix React.

Currently

  • Shipping production agents with FastAPI, tool-use loops & cost tracking
  • RAG over messy real-world documents — chunking, evals, the works
  • Treating evals like a first-class engineering discipline (because they are)
  • Wiring Next.js + React Query frontends to all of the above

Interests

  • LLM evals
  • Vector search
  • Agent frameworks
  • Developer tools
  • Design systems
  • Low-latency UIs
02 — Stack

The toolkit.

Chosen for boring reliability, not Twitter hype. Mostly.

Applied AI

  • OpenAI GPT-4o
  • Claude
  • Gemini
  • LangChain
  • LlamaIndex
  • RAG pipelines
  • Function / tool calling
  • ReAct agents
  • Evals
  • Prompt engineering
  • Embeddings
  • Whisper

Backend

  • FastAPI
  • Uvicorn
  • Pydantic
  • Node.js
  • Express
  • REST
  • WebSockets
  • Auth (JWT, OAuth)
  • Background jobs
  • uv

Frontend

  • React 19
  • Next.js
  • TypeScript
  • React Query
  • Tailwind CSS
  • shadcn/ui
  • Framer Motion
  • Vite
  • Web Speech API

Data & Storage

  • PostgreSQL
  • MongoDB
  • Redis
  • SQL
  • Pinecone
  • Chroma
  • Weaviate
  • SQLite

Cloud & Infra

  • Docker
  • Vercel
  • GCP Cloud Run
  • AWS
  • Railway
  • GitHub Pages
  • GitHub Actions
  • CI/CD

Languages & Tools

  • Python
  • TypeScript
  • JavaScript
  • SQL
  • Bash
  • Git
  • VS Code
  • Jupyter
  • Postman
  • ESLint
03 — Selected work

A few things I’ve built.

A handful of projects I’d actually defend. The rest of my GitHub is experiments and graveyards — proceed with realistic expectations.

Featured · AI TypeScript

IELTS Coach

AI-powered IELTS Academic prep covering Listening, Reading, Writing & Speaking with band-score feedback and live speech-to-text practice.

  • React 19
  • Vite
  • Tailwind
  • Gemini
  • Web Speech API
Live demo → Source
AI · Backend Python

Research Agent

Production-ready ReAct agent with tool-use, persistent memory, cost tracking and prompt-injection defense — wrapped in a FastAPI service.

  • FastAPI
  • GPT-4o-mini
  • Tavily
  • ReAct loop
Source →
RAG · Finance Python

SECfiler RAG

A retrieval-augmented pipeline over SEC filings with evals, structured tests and a clean separation between ingestion, retrieval and generation.

  • Retrieval
  • Evals
  • Chunking
  • LLM
Source →
AI · Tooling Python

LLM Extract

Schema-first structured extraction from unstructured text. Pydantic schemas, a tight LLM core, and a tiny CLI — built to slot into any pipeline.

  • Pydantic
  • Structured output
  • CLI
Source →
Backend · Production Python

Production API

A FastAPI service scaffolded the way I actually like — uv for deps, Pydantic everywhere, a real tests/ directory, clean app/test split. The boring base that lets the interesting stuff ship.

  • FastAPI
  • Pydantic
  • uv
  • pytest
Source →
AI · Lab Python

Claude AI Lab

A personal playground for prompts, tool-use patterns, and small Claude-powered agents. Where ideas earn the right to become real projects (or quietly don’t).

  • Claude
  • Agents
  • Tool use
  • Prompts
Source →
See more on GitHub
04 — Journey

How I got here.

A short timeline. Sourced from public work; intentionally light on prose.

  1. 2026 — Now

    Engineering AI agents & retrieval systems

    Shipping production agents and RAG pipelines: secfiler-rag, research-agent, LLM-Extract, production-api — focused on evals, observability and clean APIs.

  2. 2025 — 2026

    Modern web & product UI

    Built TypeScript/Next.js products end-to-end — including the IELTS Coach AI prep app and several client sites — leaning into React Query, clean motion, and typography that doesn’t scream.

  3. 2024 — 2025

    AI/ML foundations & DSA

    Worked through ML notebooks, Naive Bayes labs and a steady stream of NeetCode problems — building the algorithmic and statistical base behind everything since.

  4. December 2021

    Started on GitHub

    First commits go up. A long, slow accumulation of experiments begins — most of them learning exercises, a few that turned into the projects above.

05 — Contact

Let’s build something.

Open to full-time roles, contracts, and the rare interesting collaboration. The fastest way to my inbox is, predictably, email.

Get in touch

I read every message. Yes, including the recruiter copy-paste ones — I just reply to the others first.

rohhit.rz@gmail.com