Machine Learning Engineer
I build scalable ML and LLM systems with real-time computer vision, RAG pipelines, and autonomous agents. I deliver high-performance, cost-efficient models across the full lifecycle and contribute to open-source tools that enhance reliability and context-aware intelligence.
I build production-ready ML and LLM systems that deliver real impact, from real-time computer vision to autonomous agents and RAG pipelines. I enjoy working on challenging problems and transforming research ideas into deployable solutions.
I contribute to open-source projects, experiment with new ML architectures, and explore the latest trends in AI and software development.
Total Annual Business Savings from Efficiency & Performance Improvements
Hours of Model Training & Experimentation
Open-Source Contributions
Papers Published & Patent Granted
This project implements a lightweight OpenPose MobileNet architecture in PyTorch for real-time hand keypoint detection. Inspired by the paper “Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose”, the model leverages the Hands from Synthetic Data (CMU) dataset, which provides 21 keypoints per hand image.
This project implements a monocular 3D hand pose estimation model using a ResNet-based architecture in PyTorch. The model is trained on the FreiHAND dataset, which follows the MANO parametric hand model format, enabling seamless integration with SMPL-X or SMPL-H body models.
CustomOCR is a complete OCR training pipeline built with PyTorch and PyTorch Lightning, enabling fast and efficient dataset generation, preprocessing, and model training.
If you’d like to collaborate, share insights, or just say hi, I’d love to hear from you.