Projects

Real-Time ML Video Production Platform

Monks

Autonomous sports video production with multi-model inference

Configurable ML pipelines processing live multi-camera video streams. Chained transformer-based inference across detection, pose estimation, recognition, and segmentation — orchestrated across multi-GPU infrastructure.

  • Multi-model pipeline: vision transformers, foundation models, and VLMs for real-time analysis
  • LLM-powered autonomous video director for camera switching decisions
  • Kubernetes GPU deployment with air-gapped support and optimized model serving
PythonPyTorchKafkaKubernetesCUDATransformers

Probabilistic Time-Series Modeling

Research

Latent state models for algorithmic trading

Financial regime detection via probabilistic graphical models and sequential state inference. Applied to currency and crypto markets for adaptive strategy selection.

PythonPyTorchscikit-learnNumPy