- 🧩 Applied ML Engineer with 4+ years of experience in agentic AI, LLM systems and workflow automation
- ⚡ Focused on LLM inference: fine-tuning, quantization and high-throughput serving of open-weight models
- 🔎 Building RAG systems, AI agents and LLM evaluation for production
- 📈 Production ML: forecasting, anomaly detection and recommender systems with MLOps
| Project | Description | Tech |
|---|---|---|
| 🌐 Trellis-WebGPU | 3-bit trellis-quantized 8B LLM running a full forward pass in-browser on WebGPU — no CUDA | Python, WebGPU, WGSL |
| ⚡ Triton Kernels | High-performance GPU kernels for LLM inference (RMSNorm, SwiGLU, W4A16/W8A16 GEMMs, fused MoE) | Python, Triton, CUDA |
| 🔬 Agent Quality Inspect | User-aware evaluation of agentic AI systems: subgoal metrics (AUC, PPT, pass@k) + automated error analysis | Python, Agent Eval, LLM |
| ⏱️ Adaptive Test Time Compute | Per-instance test-time compute scaling for LLM reasoning via consensus-based adaptive sampling | Python, GSM8K, Consensus Sampling |