Abel Mendoza

Omega-1 — Autonomous Robotics Platform

A fully operational autonomous robot running on Raspberry Pi 4B — drives, navigates, streams live video, and tracks its own position in real time. Controlled from a custom web dashboard with Xbox controller support, live sensor feeds, and a mission control map.

Built as a distributed system across three machines: the Pi handles real-time motor control and sensor fusion, a web UI runs on any laptop with no robot software installed, and an NVIDIA Jetson Orin handles GPU-accelerated object detection — all communicating over a VPN-secured network.

• Drives autonomously with obstacle avoidance — ultrasonic sensor triggers deceleration, pivot, and heading recovery without stopping abruptly
• Estimates its own position using a Kalman filter that fuses dead-reckoning (motor commands) with visual landmark corrections (ArUco markers) — no GPS required
• Streams live camera video to the browser with under 100 ms latency via a custom MJPEG proxy
• Runs a full software simulation mode: the robot's physics and sensor geometry are modeled so the entire stack can be developed and tested without hardware
• Detects objects with YOLOv8 on the Jetson Orin and relays results back to the Pi in real time over a compressed image stream (~500 KB/s vs 9 MB/s raw)
• Solved a hardware bug where the motor driver corrupted servo signals at power-on — traced to a shared PWM clock; fixed with register verification and hard pulse clamping on every write
• Eliminated ~240 unnecessary HTTP requests/minute from the frontend by introducing circuit breakers, shared polling contexts, and WebSocket health checks
• Stack: Python · Go · TypeScript · React · Next.js · FastAPI · ROS2 · OpenCV · WebSockets · Docker · TailwindCSS
• Status: Fully operational — live teleoperation, sensor streaming, localization, and multi-machine communication all running.

PX4 / ROS 2 Autonomy & Verification Platform

Autonomous UAV development and verification platform built with PX4, ROS 2 Humble, Gazebo Harmonic, MAVSDK, MAVLink, Python, and C++. Features autonomous navigation, cooperative multi-UAV missions, fault injection, sensor integration, and requirements-based verification. Includes an interactive 3D flight replay application built with Three.js.

• ROS 2 mission state machines, A* path planning, LiDAR-based reactive obstacle avoidance, geofencing, failsafe handling, and resource-aware mission control
• Cooperative two-UAV missions with task allocation, route reservations, separation monitoring, watchdogs, and automatic task reassignment after confirmed vehicle failure
• Requirements-based verification: a 26-requirement catalog, independent telemetry analysis, fault injection, regression testing, and PASS/FAIL/SKIP reporting
• 450 ROS-enabled Python tests, 27 browser tests, and 41 C++ checks, with GitHub Actions CI/CD and reproducible evidence reports
• Stereo visual odometry, IMU fusion, velocity PID control, ArUco precision landing, and simulated GPS-aiding-loss handling
• MAVLink v2 communication utilities, vehicle abstraction interfaces, and telemetry pipelines for simulated and hardware-facing integration
• Two-UAV fault-recovery mission verified across 641 telemetry samples, two task reassignments, and 8.095 m minimum sampled horizontal separation against a 2.5 m requirement
• Mission-completion defects found and corrected through independent verification, with the failing evidence kept as regression tests
• Browser-based Three.js flight replay that visualizes recorded missions and fault recovery without a local PX4/Gazebo install
• Stack: PX4 v1.15+ · ROS 2 Humble · Gazebo Harmonic · Python · C++ · MAVSDK · MAVLink · Micro XRCE-DDS · OpenCV · Three.js · GitHub Actions

Jetson RealSense Perception

Embedded RGB-D perception toolkit for NVIDIA Jetson Orin Nano and Intel RealSense D455. Supports standalone pyrealsense2 workflows and ROS2 Humble integration, validated on Ubuntu 22.04 / JetPack L4T 36.4.7.

• Standalone scripts: capture, camera info probe, obstacle detection, point clouds, IMU, occupancy mapping
• ROS2 wrapper validation: realsense2_camera launch, topic sanity, camera_info intrinsics, topic publishers
• Documentation + onboarding: hardware setup, software setup, ROS2 setup, troubleshooting, validation checklist
• Full validation: rs-enumerate-devices and check_realsense.py; depth/color streaming capture; ROS2 topic and intrinsics checks

Titan Rover — Autonomous Navigation

Group project with the Titan Rover university club — a team of Cal State Fullerton engineering students. Built and contributed to an autonomous rover using ROS, OpenCV, and sensor fusion for real-world terrain navigation and obstacle avoidance.

• Autonomous terrain navigation with real-time obstacle detection using OpenCV
• Sensor fusion pipeline integrating camera, IMU, and encoder data
• ROS-based architecture with custom nodes for perception and motion planning

LoreBook — AI Life Engine

A full-stack AI platform that transforms journals, conversations, and memories into a continuously evolving knowledge base of a person's life. LoreBook automatically builds timelines, biographies, relationship intelligence, life arcs, and evidence-backed personal knowledge from unstructured entries.

• Autobiographical Memory System — converts journal entries into structured events, timelines, relationships, and life arcs that persist over time
• Knowledge Crystallization Engine — detects recurring behavioral patterns and converts them into evidence-backed knowledge claims with confidence scoring and traceable supporting evidence
• Continuity Intelligence — identifies contradictions, identity drift, abandoned goals, emotional arcs, and recurring behavioral patterns across a user's history
• Relationship Intelligence System — tracks relationship evolution, influence, interaction patterns, drift, and autobiographical impact across a user's life story
• Hybrid Retrieval Architecture — combines vector search (pgvector), BM25 keyword search, entity-aware retrieval, temporal weighting, and MMR reranking for context-aware memory retrieval
• Temporal Reasoning Engine — builds causal chains, life arcs, chronology graphs, and hierarchical timelines from years of journaled experiences
• Custom orchestration framework: 30+ specialized analysis engines with dependency-aware scheduling, parallel execution, 10–100× caching improvements, and eliminated N+1 bottlenecks
• Stack: TypeScript · React · Node.js · Supabase/PostgreSQL · pgvector · OpenAI · Stripe · GitHub Actions

CERBERUS — The Guardian Security Suite

An open-source security monitoring and system hardening ecosystem for Windows, Linux, and macOS. Three separate Guardian desktop apps — one for each platform — share one landing page: integrity checks, security posture, network exposure, threat detection, and remediation that stays dry-run until the operator confirms a change.

• Three desktop Guardians — Windows (WinUI 3 / .NET 8), Linux, and macOS — each in its own repository
• System integrity, security posture, network monitoring, and threat detection on every platform
• Controlled remediation: dry-run or confirmation-gated until the operator approves a change
• React landing page that documents what each public Guardian repository actually ships
• Stack: TypeScript · React · Vite · Tailwind CSS · .NET · Python · Shell