Course 6 · Advanced
Embodied AI — Connecting LLMs to Robotics & Physical Hardware
12 hours · 3 sessions × 4 hours · Max 10 participants
A tabletop embodied-AI inspection and sorting cell: camera and sensors → Raspberry Pi gateway → local Ollama/Qwen3-VL → deterministic safety and policy gate → ESP32 → low-voltage actuator → telemetry.
Who should attend
IoT engineers, hardware developers, robotics enthusiasts, manufacturing automation teams, and technical founders.
Where and how
12 hours · 3 × 4-hour sessions · Max 10 participants · Hands-on · English · On-site or at an agreed training premises.
What we cover
Session 1 — Hardware-to-LLM — Capture real sensor and button inputs. Package versioned JSON events. Connect to Ollama across a trusted LAN. Create strict semantic command contracts and build the policy gate between AI and hardware.
Session 2 — Visual Grounding — Capture event-driven camera frames. Use a local multimodal model for structured object interpretation. Handle UNKNOWN and confidence thresholds. Translate perception into bounded commands. Measure end-to-end latency and actuator acknowledgement.
Session 3 — Edge Agentic Loops — Build explicit state machines. Handle Wi-Fi, model, sensor, and heartbeat failure. Implement bounded retry and self-correction. Integrate the mini-MVP, run forced failure tests, and complete the technical handover.
You will learn to
- Connect sensors, cameras, local models, and microcontrollers.
- Use strict JSON contracts between models and hardware.
- Gate model proposals with deterministic safety rules.
- Build bounded edge loops with failure handling.
- Test latency, acknowledgements, and forced failures.
The safety principle
AI proposes; deterministic control disposes. The model never controls arbitrary PWM, speed, or steering, and never replaces an emergency stop, watchdog, or local safety interlock.
Investment
- B2C / open cohort: 10,000,000 VND / participant
- Private B2B cohort: 48,000,000–72,000,000 VND / private cohort
Hardware procurement, travel, and production industrial integration are quoted separately.
Completion
Attend, complete the assessment, and finish the capstone. We will give you a course completion certificate.
For private cohorts, we can adapt the examples to a cleaned-up company use case after a short technical call.