Building AI/ML systems with Python, from video retrieval and computer vision to world-model research and streaming data pipelines.
I’m interested in how models fit into complete systems: preparing data, retrieving useful context, serving predictions, and making results understandable.
-
DSpro — Local-first video search and chat using multimodal retrieval, FastAPI, and Qdrant.
-
WPMS — Computer-vision monitoring prototype combining detection, tracking, event logging, and a Streamlit dashboard.
-
Beyond Uncertainty — Ongoing research into data and model-class failures in world models, with PyTorch experiments, diagnostic infrastructure, and reproducibility records.
-
Real-Time Market Lakehouse — Streaming analytics and experimental anomaly detection with Kafka, Spark, Delta Lake, FastAPI, and a Next.js dashboard.
Each repository explains its architecture, setup, and current limitations.
- AI and retrieval: PyTorch, multimodal retrieval, Qdrant, and local language models
- Computer vision: YOLO, ByteTrack, and OpenCV
- Data systems: Kafka, PySpark, Delta Lake, and SQLite
- Applications: Python, FastAPI, Streamlit, TypeScript, and Next.js
Video retrieval and grounded question answering, practical vision pipelines, and diagnosing uncertainty in learned world models.