AI Systems Engineer focused on Physical AI, Edge AI, Agentic AI, LLM/RAG, and AI-RAN.
I build AI systems where intelligence meets infrastructure: robots, edge devices, telecom networks, distributed compute, and operational workflows.
My work sits at the intersection of:
- Physical AI and Robotics - Jetson, ROS 2, Isaac Sim, OpenUSD, sim-to-real workflows
- Edge AI Runtime Engineering - Jetson Thor/Orin, TensorRT, vLLM, VLM/LLM inference, latency and memory optimization
- Agentic AI and LLM/RAG - telemetry copilots, operational agents, retrieval systems, safety-bounded automation
- AI-RAN and Telecom AI - KPI forecasting, anomaly detection, private 5G, workload placement, network intelligence
- Production ML and MLOps - pipelines, CI, testing, experiment tracking, monitoring, explainability
I am currently pursuing an M.S. in Applied Artificial Intelligence at the University of San Diego.
A production-oriented engineering platform for building and validating AI systems that can observe, reason over, and support physical infrastructure at the edge.
The goal is not to build another notebook demo. The goal is to build a credible engineering stack for:
- robot telemetry collection and diagnostics
- Jetson edge inference and benchmarking
- ROS 2 robotics workflows
- OpenUSD and Isaac simulation
- LLM/RAG-based operations copilots
- AI-RAN and private 5G readiness for robotics workloads
- sim-to-real validation and deployment evidence
Flagship repo: physical-ai-jetson-robotics
Building simulation-to-real robotics systems around Jetson, ROS 2, OpenUSD, Isaac Sim, mobile robots, robotic arms, perception, telemetry, and edge deployment.
Repository: physical-ai-jetson-robotics
Optimizing compact AI workloads for constrained edge systems using ONNX, TensorRT, Jetson-class hardware, latency benchmarking, memory profiling, and deployment-aware evaluation.
Repository: mnist-deep-cnn-improved-image-classification
Exploring AI-assisted edge and IIoT security workflows, telemetry-aware anomaly detection, and operational observability for infrastructure environments.
Project Direction: projects/jetson-edge-ai-security/
Building interpretable AI workflows with SHAP, operational summaries, human-in-the-loop review, and agentic reporting patterns.
Repository: agentic-medical-ai-explainability
Applying AI to RAN telemetry, KPI forecasting, congestion visibility, anomaly detection, private 5G analytics, and inference-aware edge infrastructure.
Repository: ai-ran-kpi-forecasting
Designing data pipelines for network telemetry ingestion, validation, transformation, feature engineering, and downstream AI/ML workflows.
Repository: private-5g-data-pipeline
Exploring communication system foundations for AI-RAN, 6G, physical-layer intelligence, channel modeling, BER analysis, and future AI-assisted receiver workflows.
Repository: qpsk-wireless-link-simulator
| Track | Area | Repository | Focus |
|---|---|---|---|
| 1 | Physical AI and Robotics | physical-ai-jetson-robotics | Jetson robotics, ROS 2, OpenUSD, Isaac, sim-to-real |
| 2 | Edge AI Runtime and Infrastructure | mnist-deep-cnn-improved-image-classification | Edge CNN optimization and deployment benchmarking |
| 3 | Edge AI Security and Observability | projects/jetson-edge-ai-security/ |
Edge and IIoT telemetry-aware security workflows |
| 4 | Agentic and Explainable AI Systems | agentic-medical-ai-explainability | Explainable clinical decision-support workflows |
| 5 | AI-RAN and Telecom Intelligence | ai-ran-kpi-forecasting | RAN KPI forecasting and telecom operational intelligence |
| 6 | Telecom Telemetry and Observability | private-5g-data-pipeline | Network telemetry pipelines and observability infrastructure |
| 7 | Wireless Systems and Signal Processing | qpsk-wireless-link-simulator | QPSK simulation, BER analysis, wireless link foundations |
Supporting repo:
| Area | Repository | Focus |
|---|---|---|
| Telecom Customer Intelligence | telecom-churn-ml-with-agents | Customer intelligence, explainability, and agentic analytics |
These are planned project directions that support the same core thesis of deployable AI systems for edge and physical infrastructure:
| Direction | Focus |
|---|---|
| Physical AI Safety Observability | Cosmos/VLM-style safety monitoring, PPE verification, and industrial observability |
| Urban Edge Vision Analytics | Human-reviewed traffic event analysis and smart intersection edge intelligence |
Project notes live under projects/.
I am shaping this GitHub into a focused portfolio around one thesis:
Production AI is moving from isolated models to deployed systems that reason over physical infrastructure, edge compute, network telemetry, and real-world operations.
That means the work here increasingly focuses on:
- robotics and Physical AI systems
- edge inference on Jetson-class hardware
- agentic operations copilots
- RAG over logs, manuals, telemetry, and system state
- AI-native telecom and AI-RAN workflows
- explainable and human-reviewed AI systems
- reproducible engineering, testing, and deployment evidence
AI and ML: Python, PyTorch, scikit-learn, XGBoost, SHAP, MLflow
LLM and Agentic Systems: RAG, tool-using agents, guardrails, local inference workflows
Edge AI: NVIDIA Jetson, TensorRT, vLLM, ONNX, VLM/LLM deployment
Robotics: ROS 2, MoveIt 2, Isaac Sim, Isaac Lab, OpenUSD, sensor-driven systems
Data and MLOps: SQL, Spark, Airflow, dbt, Docker, Kubernetes, CI/CD
Cloud and Infrastructure: AWS, Azure, GCP, Terraform, distributed compute
Telecom and AI-RAN: RAN telemetry, KPI forecasting, private 5G, edge workload intelligence
- Email: obiedeh@gmail.com
- LinkedIn: linkedin.com/in/obinna-edeh-206306137
- GitHub: github.com/obiedeh