Vision stack for autonomous drones to identify targets, map terrain, and make decisions without human intervention.
Context: Competing globally to build autonomous planes for search-and-rescue missions.
Stack: C++, YOLO v11, ONNX Runtime, OpenCV, Jetson Orin Nano, ArduPilot SITL, Python
Result: Real-time aerial target detection and geolocation with synthetic data validation pipeline.
Learn More →A native extension for Adobe Premiere Pro to randomize clip order, solving a major workflow bottleneck.
Context: Video editors had no native way to randomize clips in Adobe Premiere Pro.
Stack: Adobe CEP, ExtendScript, HTML, CSS
Result: 8,000+ users, 4.9-star rating on Adobe Exchange.
Learn More →A personalized RAG assistant that ingests notes and conversations to act as a "Second Brain" with perfect recall.
Context: Addressing the limitations of human memory and context retention.
Stack: FastAPI, LangChain, Pinecone, Docker, Streamlit, Supabase
Result: AI assistant with high-recall personal memory using time-weighted RAG and preference learning.
Learn More →Comparative benchmarking of LSTMs vs. Transformers for fake news detection, highlighting generalization gaps.
Context: Research on deep learning robustness.
Stack: PyTorch, LSTM, GloVe, DistilBERT, GPT-4 Turbo
Result: 99.43% accuracy on in-distribution data, exposed generalization gap in pre-LLM models.
Learn More →
