I'm a Fullstack Engineer with 3+ years of experience building production-grade software at the intersection of AI and backend infrastructure.

Right now, I'm fine-tuning Qwen3.6-35B-A3B (a 35-billion-parameter large language model) on AMD MI300X GPUs. The model achieved zero refusals across 465 safety tests using a technique called heretic-abliteration, with safety policies enforced at the application layer.

I optimize for latency, cost, and correctness, without the shortcuts that bite you 6 months later.

Languages
AI / ML
Backend
Frontend
Databases
Cloud / DevOps
Tools

Domain-specialized AI fine-tuning project. Built for the AMD Hackathon 2026. Specialized fine-tune of Qwen3.6-35B-A3B (a 35-billion-parameter large language model) on AMD MI300X GPUs. Achieved zero refusals across 465 safety tests via a technique called heretic-abliteration, with safety policies enforced at the application layer.

Open-source AI research project. Research on safe model abliteration and post-training quantization (GPTQ, AWQ, and GGUF formats). Includes tooling, evaluations, and reproducible recipes for the broader community.

Production-grade agentic AI chatbot. Built with typed tool calls, persistent memory, and streaming responses. Provider-agnostic, so it works seamlessly with multiple AI providers (OpenRouter, Ollama, and Anthropic).

Real-time disease tracking dashboard. Pulls live data from the World Health Organization (WHO) and Johns Hopkins University. Features map-based geospatial visualization and trend forecasting.

I'm currently open to senior or staff engineer roles (individual contributor positions, not management).

I'm looking for opportunities in:

• 🤖 Applied AI / LLM Infrastructure · fine-tuning, inference, agentic platforms
• ⚙️ Backend Platforms · high-throughput Python or Java services, distributed systems
• 🛠 Developer Tooling · agentic systems, LLM-powered developer tools

I'm based in India and open to remote roles or relocation. I respond to all messages within 24 hours.