Shushant Kamatar

Machine Learning Engineer with 1.5+ years of experience building scalable, production-grade AI solutions. Unlike researchers who stop at the Jupyter Notebook, I specialize in End-to-End Machine Learning Engineering. I take raw data and turn it into deployed, low-latency APIs that drive business value. My Core Competencies: Full-Stack ML: Proficient in the entire lifecycle—from data ingestion (SQL/Pandas) to model training (Scikit-Learn/PyTorch) and deployment (FastAPI/Docker). GenAI & NLP: Architected RAG pipelines and low-latency inference engines using modern stacks (Llama 3, LangChain, Vector DBs). System Reliability: Proven track record at Lenovo optimizing high-volume data pipelines, reducing processing time by 99% and ensuring system stability across 15 markets. Technical Stack: Languages: Python, SQL, JavaScript. ML & AI: Scikit-Learn, NLP, Transformers, RAG, Computer Vision. MLOps & Cloud: Docker, Kubernetes, AWS, GitHub Actions, MLflow. Backend: FastAPI, AsyncIO, Redis, PostgreSQL. I am methodology-agnostic: I use whatever tool solves the problem efficiently—whether that's a simple Regression model or a complex Agentic Workflow.

Machine Learning Engineer

Confidential (Freelance) Nov 2025 - Present

• Architected a real-time Interview Intelligence System using FastAPI and WebSockets, achieving 800ms latency by parallelizing audio transcription (Deepgram) and LLM inference (Llama 3).
• Engineered a context-aware RAG pipeline with Supabase (pgvector) to dynamically retrieve candidate resume data and historical "winning questions" during live calls.
• Developed a secure Chrome Extension integration using the 'Offscreen' API pattern to capture high-fidelity audio streams from Google Meet while bypassing browser sandbox restrictions.

Software Developer

Lenovo May 2024 - Sept 2025

• Engineered a full-stack sentiment analysis dashboard for Lenovo products, integrating a Python backend with Flask for a dynamic frontend and live NLP results.
• Automated critical file merging workflow with Python and pandas, reducing runtime from 4 hours to 30 seconds and significantly boosting team productivity.
• Spearheaded rollouts and project implementations across Western Europe, ensuring successful deployment and user adoption in 15 countries.

B.E in Aerospace Engineering from RV College of Engineering