name: Harsha G M
role: Robotics and AI Systems Engineer
focus:
- Autonomous Mobile Robots (AMR)
- Edge AI and Real-time Perception
- Full-stack Robotics (Hardware to Software)
- LLM-powered Intelligent Systems
stack:
- ROS2 Jazzy | Nav2 | SLAM Toolbox
- Python | C++ | C | Embedded Systems
- FastAPI | LangGraph | RAG Pipelines
- Docker | Raspberry Pi | ESP32
currently: Seeking Robotics Engineer roles in Bengaluru
philosophy: "Ship robots that work in the real world, not just in simulation."| Domain | What I Do | Stack |
|---|---|---|
| 🤖 Autonomous Navigation | SLAM mapping, path planning, obstacle avoidance on real hardware | ROS2, Nav2, AMCL, LiDAR |
| 🧠 Edge AI and Vision | Real-time object detection optimized for embedded devices | YOLOv8, OpenCV, RPi/Jetson |
| 📡 Sensor Fusion | Multi-sensor EKF fusion, odometry calibration, state estimation | robot_localization, IMU, Encoders |
| ⚡ Embedded Systems | Motor control firmware, custom UART protocols, PID loops | ESP32, FreeRTOS, C++ |
| 🧬 AI / LLM Systems | RAG pipelines, multi-agent workflows, semantic search | LangGraph, FastAPI, Qdrant |
| 🛡️ Systems Engineering | eBPF telemetry, Kafka streaming, real-time monitoring | C, InfluxDB, Grafana |
|
Full ROS2 autonomous navigation stack — SLAM mapping, Nav2 path planning, differential-drive odometry, and custom UART hardware interface on Raspberry Pi 5 + ESP32.
|
Multi-sensor EKF fusion benchmark on the U-Michigan NCLT dataset — fuses degraded wheel odometry with IMU via
|
|
Real-time object detection and perception pipeline using YOLOv8 optimized for edge devices (Raspberry Pi / Jetson).
|
RAG-based hardware engineering copilot — hybrid vector retrieval, LangGraph state machines, Redis semantic caching, and automated RAGAS evaluation pipeline.
|
|
eBPF-powered security telemetry pipeline — captures kernel-level execve syscalls, streams via Apache Kafka, stores in InfluxDB, visualizes in Grafana with Telegram alerting.
|
Production-grade Multimodal RAG pipeline — ingests PDFs, images and tables with ColPali vision embeddings, Qdrant vector search, and citation-tracked FastAPI serving.
|
My flagship project: a full autonomous navigation pipeline running on real hardware.
RPLiDAR A1M8 --> SLAM Mapping --> Nav2 Planning --> Motor Control --> ESP32 Actuators
^ |
+---------------------- Encoder Odometry <-- 50 Hz PID Loop <--------+
+-------------------------------------------------------------+
| Raspberry Pi 5 -- ROS2 Jazzy |
| +--------------+ +--------------+ +------------------+ |
| | AMCL | | NavFn | | DWB Controller | |
| | Localization|->| Global Plan |->| Local Planner | |
| +------+-------+ +--------------+ +--------+---------+ |
| | | |
| /tf: map->odom /cmd_vel |
| ^ | |
| +------+-------+ +---------v----------+ |
| | kali_base |<--- UART -------->| ESP32 (FreeRTOS) | |
| | Odometry | Binary Protocol| 50 Hz PI Control | |
| +--------------+ +--------------------+ |
+-------------------------------------------------------------+
+ Building production-grade AMR navigation stacks with ROS2
+ Exploring multi-modal RAG and agentic AI architectures
+ Designing custom UART protocols for real-time embedded control
+ Benchmarking sensor fusion algorithms on real-world datasets
! Learning: Gazebo simulation, IMU fusion via EKF, fleet orchestration
# Open to: Robotics Engineer roles | AMR fleets | Edge AI systems
