AI Architect and Systems Engineer focused on Physical AI, Edge Inference, Runtime Observability, and AI Infrastructure.
I build AI systems where models meet physical infrastructure: robots, Jetson-class edge devices, runtime telemetry, safety-aware observability, and operator-assist intelligence workflows.
Production AI is moving from isolated models to deployed systems that reason over physical infrastructure, edge compute, telemetry, and real-world operations.
I am currently pursuing an M.S. in Applied Artificial Intelligence at the University of San Diego, with a portfolio focused on Physical AI, edge runtime systems, operational observability, and infrastructure-aware AI deployment.
My work is converging around one systems problem:
physical infrastructure
+ edge inference
+ runtime telemetry
+ operational observability
+ retrieval-grounded intelligence
+ human-in-the-loop review
= deployable AI systems for real-world environments
The focus is not generic AI demos.
The focus is building AI systems that can be:
- benchmarked
- observed
- validated
- deployed
- audited
- operated safely under real runtime constraints
This portfolio is built to prove:
- edge inference under real runtime constraints
- telemetry-driven validation and observability
- safety-aware Physical AI workflows
- reproducible engineering artifacts
- human-in-the-loop operational review
- infrastructure-aware AI deployment
| Priority | Track | Repository | System Focus | Evidence Status |
|---|---|---|---|---|
| 1 | Physical AI / Robotics Systems | physical-ai-jetson-robotics | Jetson-class robotics platform for ROS 2 workflows, robot telemetry, edge inference, and sim-to-real evidence | Active flagship |
| 2 | Physical AI Safety and Observability | physical-ai-safety-observability | Safety-observability runtime for telemetry ingestion, threshold monitoring, incident review, and evidence chains | Runnable scaffold |
| 3 | Edge AI Runtime Security | jetson-edge-ai-security | Defensive edge runtime for telemetry parsing, anomaly detection, alerting, and deployment reports | MVP runtime |
| Priority | Track | Repository | System Focus | Evidence Status |
|---|---|---|---|---|
| 4 | Urban Edge Vision Intelligence | urban-edge-vision-analytics | Edge vision workflow for frame analysis, infrastructure events, operator summaries, and deployment planning | Mock path validated |
| 5 | Private 5G Telemetry Infrastructure | private-5g-data-pipeline | Supporting telemetry pipeline for KPI ingestion, validation, feature generation, and infrastructure reporting | Fresh run pending |
| 6 | AI-RAN Operational Intelligence | ai-ran-kpi-forecasting | AI infrastructure bridge for KPI forecasting, congestion signals, and operational network intelligence | Operational report pending |
| 7 | Wireless Link Intelligence | qpsk-wireless-link-simulator | Foundational wireless simulator for QPSK behavior, BER/SNR sweeps, and link-estimation experiments | Sweep report pending |
| Priority | Track | Repository | System Focus | Evidence Status |
|---|---|---|---|---|
| 8 | Foundational CNN Optimization | mnist-deep-cnn-improved-image-classification | Foundational CNN optimization for training, evaluation, and future ONNX/TensorRT edge deployment practice | Foundation project |
| 9 | Telecom Agentic Analytics | telecom-churn-ml-with-agents | Foundational agentic workflow for churn risk, explainability, and human-reviewed customer intelligence | Evaluation pending |
| 10 | Explainable Human-Reviewed AI | agentic-medical-ai-explainability | Foundational explainability workflow with SHAP, safety boundaries, and reproducible human-review reports | Safety caveats pending |
Some projects live as standalone repositories; the projects/ folder contains selected profile-linked notes and mirrors.
Repository: physical-ai-jetson-robotics
A Physical AI engineering platform connecting:
- ROS 2 robotics workflows
- Jetson edge inference
- OpenUSD / Isaac simulation
- robot telemetry and diagnostics
- sim-to-real validation
- safety-aware operations tooling
- retrieval-grounded diagnostics over logs, documentation, and runtime state
- AI-RAN / private 5G readiness concepts for robotics workloads
This repository is the center of gravity for the portfolio.
Physical infrastructure
-> edge inference
-> runtime telemetry
-> operational observability
-> retrieval-grounded intelligence
-> operator-assist workflows
-> telecom / wireless infrastructure support
The repositories are intentionally connected.
The broader thesis is that AI systems become operationally valuable only when connected to:
- telemetry
- runtime constraints
- evidence
- observability
- human review
- deployment workflows
The current portfolio focus is strengthening the flagship proof stack:
physical-ai-jetson-robotics— Jetson/runtime evidence, telemetry artifacts, sim-to-real validationphysical-ai-safety-observability— safety event evidence, operator review flow, runtime metricsjetson-edge-ai-security— defensive telemetry replay, alert artifacts, runtime reporting
Detailed maturity tracking is maintained in PORTFOLIO_EVIDENCE.md.
Hiring-manager mapping: HIRING_MANAGER_BRIEF.md.
Each flagship project should include:
- reproducible run command
- tests / CI validation
- runtime metrics artifact
- architecture diagram
- sample input/output
- limitations
- next validation step
This portfolio separates:
- implemented workflows
- runnable scaffolds
- mock validation paths
- planned hardware benchmarks
- future deployment targets
Project READMEs should state limitations clearly. Mock adapters, synthetic inputs, and planned Jetson paths are useful engineering scaffolds, but they are not claimed as real-world deployment proof until committed evidence artifacts exist.
This profile repository also includes agent operating standards for AI-assisted development:
AGENTS.md— repository-level operating contract for Claude Code, Codex, Cursor, Aider, and similar coding agentsagent-skills/— review skills for architecture, runtime stability, observability, edge deployment, AI-RAN workflows, RAG/telemetry copilots, and sim-to-real validation
The goal is to keep AI-assisted development disciplined:
- small patches
- explicit scope
- measurable validation
- operational realism
- evidence-backed claims
- strict public/private boundaries
Edge AI: NVIDIA Jetson, TensorRT, vLLM, ONNX, CUDA, VLM/LLM deployment
Robotics / Physical AI: ROS 2, MoveIt 2, Isaac Sim, Isaac Lab, OpenUSD
AI / ML: Python, PyTorch, scikit-learn, XGBoost, SHAP, MLflow
Operational AI / RAG: retrieval-grounded copilots, local inference workflows, guardrails, human review
Data / Infrastructure: SQL, Spark, Airflow, dbt, Docker, Kubernetes, CI/CD
Telecom / AI-RAN: RAN telemetry, KPI forecasting, private 5G, wireless link analysis
Cloud / Distributed Systems: AWS, Azure, GCP, Terraform
- Email: obiedeh@gmail.com
- LinkedIn: linkedin.com/in/obinna-edeh-206306137
- GitHub: github.com/obiedeh