I'm Chinmay Rozekar — I build AI-driven automation and validation tooling for semiconductor and high-performance computing systems.
6+ years across AMD (Ryzen SoC validation, yield improvement from 42% → 85%) and Siemens EDA (Calibre PERC, DRC/LVS regression automation). Currently applying LLMs, RAG, and agentic AI to the hardest unsolved problem in chip development: making post-silicon failure triage fast.
MS in Electrical Engineering — Rochester Institute of Technology.
|
triagent — Production-ready agentic RAG platform for large-scale log triage Drain3 template mining · FAISS semantic search · Ollama local LLMs · parallel log parsing · documentation-grounded root-cause reports. Designed for semiconductor and systems-debug environments. Runs entirely on-prem — no data leaves the lab. 📄 Preprint: High-Throughput Log Parsing Architectures for Semiconductor and Network Environments 
|
| medical-rag-assistant |
RAG-based medical AI assistant using Mistral-7B, SentenceTransformers, and ChromaDB over a 4,000+ page medical manual. LLM-as-judge evaluation shows superior performance over baseline models on hallucination rate. → repo |
| superkart-sales-forecasting |
MLOps pipeline: Flask REST API + Streamlit frontend + Docker, deployed on Hugging Face Spaces. Real-time forecasts over 8,763 transaction records supporting quarterly inventory planning. → repo |
| helmnet-helmet-detection |
Computer vision safety system using VGG-16 transfer learning on 631 workplace images. Automated helmet detection for construction and industrial compliance. → repo |
| Predictive Analytics Portfolio | Gradient Boosting, Decision Trees, and DNNs with SMOTE across 40,000+ records (loan, visa, churn). Up to 99.3% recall and 81.1% F1-score across financial and immigration datasets. |
AI / ML
Languages & Automation
Infrastructure & MLOps
Open to AI/ML infrastructure, GPU validation tooling, and agentic systems roles — especially where Python automation meets large-scale hardware or model deployment.