██████╗ ███████╗██╗██╗ ██╗ ██████╗
██╔══██╗██╔════╝██║╚██╗ ██╔╝██╔═══██╗
██████╔╝█████╗ ██║ ╚████╔╝ ██║ ██║
██╔══██╗██╔══╝ ██║ ╚██╔╝ ██║ ██║
██║ ██║███████╗██║ ██║ ╚██████╔╝
╚═╝ ╚═╝╚══════╝╚═╝ ╚═╝ ╚═════╝
Building intelligent systems that understand and interact with the world
ML Engineer at Synexian Labs (New Jersey, USA) focused on computer vision, deep learning, and 3D graphics. I build systems that go from research prototype to production — with a particular interest in pose estimation, graph neural networks, and real-time visualization.
Currently working on:
- GTransformer — Graph Transformer architectures for human pose estimation
- MocapViewer3D — Real-time 3D motion capture visualization tools
- Topic-Modeled Curriculum Learning — Smarter training strategies for neural networks
ML / Deep Learning
Languages
Infrastructure
| Project | Description | Stack |
|---|---|---|
| GTransformer | Graph Transformer for human pose estimation using skeletal GNNs | PyTorch · GNN · Transformers |
| MocapViewer3D | Real-time 3D/2D motion capture viewer with multi-perspective rendering | Python · OpenGL · OpenCV |
| 2DPoseEstimation | End-to-end pose estimation pipeline with real-time inference | PyTorch · OpenCV |
| Topic-Modeled CL | Curriculum learning via topic modeling for better NN training | TensorFlow · NLP |
| CoreML-Keras3 | Keras 3.x → CoreML conversion pipeline for iOS deployment | Keras · CoreML |
| Golfbot WizardLM | LangChain agent with golf-domain tools — rules, stats, trajectory | Ollama · LangChain |
| Model | Accuracy | F1 | Status |
|---|---|---|---|
| GTransformer-v3 | 95.8% | 0.961 | ✅ Deployed |
| PoseNet-Enhanced | 93.2% | 0.945 | 🔄 Training |
| Vision-RL-Agent | 89.5% | 0.902 | 🧪 Experimental |
| BaselineNet (ResNet-50) | 87.3% | 0.888 | 📊 Baseline |
Python 12 hrs 45 mins ████████████░░░░░░░░ 55.2%
C++ 4 hrs 32 mins ████░░░░░░░░░░░░░░░░ 19.7%
Jupyter 3 hrs 15 mins ███░░░░░░░░░░░░░░░░░ 14.1%
Markdown 1 hr 23 mins █░░░░░░░░░░░░░░░░░░░ 6.0%
Other 1 hr 10 mins █░░░░░░░░░░░░░░░░░░░ 5.0%
- ❌ Closed PR #4 in RyoK3N/Synexcript
Open to: Research Collaboration · Open Source · ML Engineering Roles · Speaking
"The best way to predict the future is to invent it." — Alan Kay

