A small selection of things I’ve built, explored, or written about. Use the category bar to filter by theme, and the date tag to see when each project was published or completed.
LeX-O
Hierarchical robot agent combining Vision-Language-Action (VLA) models for board perception and planning with Action Chunking Transformer (ACT) for multi-task pick-and-place control across 9 grid positions. Built for LeRobot Hackathon using task instruction embeddings for conditional imitation learning.
Robotics Foundation Models
Comprehensive analysis of Vision-Language-Action (VLA) models, RT-1/RT-2 architectures, and the shift from classical robotics to embodied AI. Explores end-to-end learning from pixels-to-actions, semantic grounding, and cross-task generalization in physical AI systems.
ConsumerWise
End-to-end OCR + LLM pipeline using GPT-4o Vision API for nutritional label extraction, FastAPI backend for context-aware NutriBot chatbot, and interactive data visualization. Built with Flask, deployed on Google Cloud App Engine & Cloud SQL with personalized recommendations based on user dietary preferences.
Route Optimizer
Chrome extension solving Traveling Salesman Problem (TSP) via Google OR-Tools solver with support for one-way and round-trip routes. Frontend extracts locations from Google Maps; FastAPI backend optimizes order; Nominatim geocoding & OSRM distance matrix for coordinate conversion and routing.