AI Engineer India
I build LLM applications, RAG systems and voice agents, and the tooling that keeps them reliable in production.
focus LLM apps · RAG · agents and tool use · evaluation and observability
building voice agents that hold real phone conversations
learning LLMOps · distributed training · system design
off-hours chess (queensGAMBIT) · football
| Project | What it is | Stack |
|---|---|---|
| AgentArgus | Evaluation, observability and reliability for LLM agents, framework-agnostic. RAG and agent-behaviour metrics, OpenTelemetry traces, retries, fallbacks and circuit breakers. Published on PyPI. | Python · OpenTelemetry |
| CallFlow-AI | A 24×7 AI calling desk: dials contacts, holds real conversations, extracts typed results and escalates only what needs a person. Built for the CALL-E hackathon. | FastAPI · Next.js |
| RAAAAG | Voice-enabled, cross-lingual RAG over 14 Indic languages, with cited answers and per-stage latency. Built for Hacker House Goa. | FAISS · ONNX · FastAPI · Next.js |
| pyoverload | Function and method overloading for Python based on type signatures. Published on PyPI as methodoverload. |
Python |
| LaTeX OCR | Turns photos of equations into LaTeX with Llama 3.2 Vision, fully offline. | Ollama · Streamlit |
| Kokoro | Cafe ordering web app and PWA with live order tracking and a role-based admin dashboard. | Next.js · Supabase |
AI / ML
Backend & infra
Frontend & mobile
I'm learning in public. Notes on RAG, LLM inference and system design are coming soon.
Open to interesting AI engineering problems. Say hi at arbaazcode@gmail.com.