I'm an ML / Edge-AI engineer with an MS in Computer Engineering from RIT (Kate Gleason College of Engineering) who likes a good challenge: making capable models actually run under real-world memory, latency, and power budgets.
- 🧠 Research focus: neuromorphic & edge AI - spiking neural networks, RWKV/linear-attention, model quantization & sparsity, on-device deployment (BrainChip Akida, NVIDIA Jetson).
- 🚗 Building: a Vision-RWKV autonomous-driving stack (CARLA → edge) for my MS capstone, and a production LLM multi-agent platform for robot-robot interaction research.
- 🔬 NSF AWARE-AI NRT Trainee · ex-Machine Learning Engineer Intern @ BrainChip (neuromorphic semantic segmentation).
- 🛠️ I also self-host my own ML infrastructure - a multi-node WSL2 + Tailscale homelab, an n8n+Gemini agentic assistant, and a Nextcloud private cloud.
- ⚡ Interests across the stack: LLM orchestration/agents, quantum ML, robotics (ROS2), embedded/real-time systems, and HPC/CUDA.
| Project | What it is | Stack |
|---|---|---|
| 🧠🚗 neuromorphic-vision-rwkv | MS capstone - an RWKV linear-attention autonomous-driving model trained on a 3.6 TB CARLA corpus; 6.88 ms median inference, 13.4 M params, targeting Akida/Jetson edge deployment. | PyTorch CARLA RWKV Edge AI |
| 🤖💬 rri_orchestrator | Production-deployed LLM multi-agent platform for robot-robot interaction research - async FastAPI + NiceGUI, PostgreSQL, LiteLLM (GPT-4o/Gemini/Claude), RBAC, batch automation, self-hosted via Cloudflare Tunnel. | FastAPI LiteLLM PostgreSQL Docker |
| ⚡🖥️ VGG16_Optimization | Custom CUDA C++ kernels for VGG-16 inference (shared-memory tiling, memory coalescing, cuBLAS); Nsight Compute roofline/CGMA analysis pinned the FC layers memory-bound (~41.5 GFLOPS @ batch=16). 6-person team. | CUDA cuBLAS Nsight Compute C++ |
| ⚛️🎮 Quantum_Policy_Gradient_for_CartPole | A 4-qubit variational quantum circuit RL agent benchmarked head-to-head against a parameter-matched classical net - statistical parity at 18% fewer params, validated with the hardware-compatible parameter-shift rule. | PennyLane PyTorch Gymnasium |
💡 More context on my work, research, and experience: linkedin.com/in/shahzebkjadoon
Languages
ML / Deep Learning
Edge AI / Neuromorphic
LLM / Agentic AI
Robotics / Autonomy / Embedded
Quantum ML
Infra / DevOps
⚡ "Train dense, deploy sparse." - currently teaching big models to live on small chips.