I build autonomous navigation stacks for ground rovers (SLAM, localization, path planning) and the embedded/edge AI systems that go with them — from sensor input to motor output.
⚙️ Available for Freelance & Contract Work in robotics hardware/software architecture, embedded systems, and machine vision.
🛠️ Autonomy, Navigation & Drive Systems
- ROS2 Navigation Stacks: Complete integration of SLAM (
FAST-LIO,SAM-LIO,Cartographer,slam_toolbox,ORB-SLAM), point-cloud localization (NDT scan matching,GICP,TEASER/TEASER++), path planning (Nav2,pure pursuit), and precision PID control tuning. - Drive Kinematics: Custom drive system configurations for Ackermann steering, 2WD/4WD skid-steer, and Mecanum-wheel platforms. Experienced with Clearpath rovers and TurtleBot ecosystems.
- Simulation & Point-Clouds: Virtual environment configurations via
NVIDIA Isaac SimandGazebo; post-processing withRTAB-MapandCloudCompare.
🔌 Embedded Firmware & Sensor Integration
- Sensors: Dense/sparse LiDAR arrays (
Livox MID-360,RPLiDAR,ST/STM32 LDx series), stereoscopic/depth vision (ZED,RealSense), monocular depth estimation network deployment,mmWave&UWBspatial ranging. - Firmware: Custom
ESP32(includingESP32-CSItracking) andSTM32(STM32CubeIDE/CubeMX) code bases written for low-latency motor control, custom sensor telemetry parsing, and radio-link configurations.
👁️ Edge AI, Drones & Rapid Hardware Prototyping
- Computer Vision: End-to-end perception pipelines (object detection, spatial tracking, feature recognition) compiled into edge-friendly runtimes (
TensorFlow Lite,ONNX) for deployment onNvidia JetsonorRaspberry Pi. - Aviation:
ArduPilot-based flight stack setups, payload mechanics, customized motor/ESC dynamometer test benches, and structural integration for specialized delivery and agricultural drones. - Hardware Integration: Speed-optimized breadboard/perfboard electrical prototyping, custom battery management system (BMS) logic, and robust
FDM 3D-printedcustom structural enclosures. Proven runtime optimization: going from raw idea to fully functional hardware prototype inside 24 hours.
Global 3D LiDAR Relocalization package for ROS2 Humble.
- Recovers global coordinates without a prior pose estimate by executing a robust correspondence-free global registration (
TEASER++) combined with sequential iterative closest point adjustment (GICP). - Plugs directly into standard
FAST-LIOpipelines and broadcasts correct coordinate updates toNav2via proper TF fusion. - Built to handle sparse, non-uniform spatial scans (e.g.,
Livox MID-360) that typical naive point-matching frameworks fail to resolve.
Fully autonomous micro-rover turned around from concept to operating hardware in under 15 hours.
- Integrates time-of-flight (ToF) obstacle collision avoidance and publishes live IMU/LiDAR telemetry matrices via concurrent
WebSocketandUDPstreams. - Provides an instantly accessible browser-based mobile controller dashboard requiring zero client app installation.
An end-to-end computer vision tracking system optimized for live processing.
- Built from raw framework layer up: proprietary training-set capture routines, structural dataset data augmentation, and a dual-headed
VGG16network topology configured for concurrent classification and coordinate bounding-box regression. - Utilizes a purely hand-written custom Keras training loop running dynamic inference pipelines live over raw webcam streams.
Production-ready standalone spatial mapping and odometry toolkit.
- Features a proprietary occupancy-grid calculation pipeline designed to drop strict structural TF-tree constraints typically requested by standard
slam_toolbox. - Includes efficient
PCL-to-LaserScanconversion utilities and a structural pose estimation model trained directly on filtered, downsampled raw LiDAR scans.
High-performance minimalist vector database engine engineered from scratch in native C++.
- Zero massive external framework dependencies; includes an in-house engineered word embedding model optimized for memory constraints. Designed for applications where sub-framework compute control is required.
| 🤖 Autonomy & Controls |
• ros2_ackermann_path_planner — Hybrid A* path planning for car-like chassis. • ros2_ackermann_to_motor — Translation engine parsing commands to hardware. • ros2_ackermann_motor_ctrl — Serial motor bridge running over Arduino Mega. • ros2_ackermann_keyboard_teleop — Custom keyboard validation layouts. • ROS2_pcd_to_occupancy_grid — Automated point-cloud projection engine. • 3d-mapping-rover — Portable mapping rig using FAST-LIO & Livox hardware. • wormhole_nav — Multi-map transit transitions tracking via SQL data engines. • realsense_LFR_ackermann — RealSense vision-based line tracking for AGVs. • Qt_turtlebot_relocalizer — Performance PyQt console dashboard tracking odometry. • Autnomous-nav-DOCS — Core autonomy engineering notes & setup scripts. |
| 💻 Vision & Embeds |
• openCV_projects_cpp / openCV_projects_python — Low-latency computer vision routines. • arduino-files — Custom firmware scripts for BLDC/ESC benches, drone lift structures, and RC conversions. • edgeAI-cctv — Capstone deployment: Low-power edge neural network inferencing over CCTV stream grids. |
| 🧠 Core Foundations |
• neural-network — Forward/backward mathematical propagation engines in raw matrices. • simple-XOR-perceptron — Fundamental classification mathematical checks. • text_completion_NLP — Minimal NLP word matrix sequencer. • pyTorch-cpp-Tutorials — High-performance execution configurations using libtorch C++ bindings. • Verilog MIPS Processor — Hardware descriptive logic verification for computing cores. |
| ⚡ Other Projects |
• water_suply_management — Flask structural tracker back-ended via SQLite. • phonebook — High-efficiency trie-based lookup tracking array structure. • code_templates / dot-config — My Arch Linux daily driver setup environment dots. • Rust_simple / c-tutorial — Language specific syntax engineering trackers. |
- Robotics & Control Autonomy: Nav2 Architecture • Pure Pursuit Mathematics • PID Gain Tuning Loops • Kinematic Matrices (Ackermann / Skid-Steer / Mecanum Platform Drives) • Clearpath Systems • AMRs & UAV Airframe Layouts • FDM Structural 3D Printing Production.
- SLAM & State Estimation: FAST-LIO / SAM-LIO Solid-State & Mechanical LiDAR Pipelines • Cartographer & slam_toolbox Map Tuning • NDT / GICP Point Registration Engines • TEASER++ Arbitrary Frame Alignments • Dense Cloud Diagnostics (
CloudCompare). - Perception Array Ecosystems: Livox Solid-State Array Systems • RPLiDAR Scanners • STM32 Custom Serial Ranging Engines • ZED & Intel RealSense Stereoscopic Depth Perception Units • FMCW mmWave Radar Sensors • Ultra-Wideband (UWB) Spatial Positioning Grids.
- Embedded Hardware Engineering: Custom ESP-IDF Systems • STM32 HAL & Low-Layer Low-latency Execution Pipelines • Raw Hardware Transceiver Bring-Up (FTDI, Logic Analyzers, UART/SPI/I2C diagnostics) • Custom Prototyping Power Configurations & Battery Management Units.
- Simulation, Graphics & Data Engines: Nvidia Isaac Sim Systems • Gazebo Classic/Harmonic Ecosystems • PyQt Desktop Control Engine Engineering • Solid SQL/NoSQL Infrastructure Arrays (
PostgreSQL,MongoDB).