Open Source · Defense Tech · Applied AI

Engineering systems
at the edge of autonomy

An independent R&D lab building open-source defense technology, autonomous systems, and edge AI. We operate at the intersection of software, hardware, and contested environments.

12+
Open Source Projects
8
Research Publications
3K+
GitHub Stars
15+
Years in R&D

Where software meets the physical world

Bridging the gap between cutting-edge research and deployable systems. Every project starts with a hypothesis, ends with working code.

Autonomous Drone Systems

Flight controllers, SLAM-based navigation, swarm coordination protocols, and edge compute payloads for BVLOS operations.

🧠

Edge AI & Inference

Real-time neural network inference on constrained hardware. ONNX, TensorRT, and custom quantization for NVIDIA Jetson and FPGA targets.

📡

Defense & ISR Software

C2 systems, sensor fusion pipelines, EW signal processing, and tactical mesh networking for contested environments.

🔎

Computer Vision & GEOINT

Object detection, change detection, SAR imagery analysis, and multi-spectral classification for aerial and satellite feeds.

🔗

Open Source Tooling

Developer tools, SDKs, and frameworks released under permissive licenses. Building in public, shipping fast, iterating with the community.

🔐

Offensive Security Research

Vulnerability research, embedded systems security, firmware reverse engineering, and red team tooling for critical infrastructure.

Rust
C++
Python
ROS2
PX4
ArduPilot
NVIDIA Jetson
PyTorch
ONNX
TensorRT
OpenCV
gRPC
MQTT
Zig
Linux Kernel
FPGA / Verilog
Rust
C++
Python
ROS2
PX4
ArduPilot
NVIDIA Jetson
PyTorch
ONNX
TensorRT
OpenCV
gRPC
MQTT
Zig
Linux Kernel
FPGA / Verilog

From hypothesis to deployment

Rigorous engineering process. Every system is designed to survive contact with the real world.

01

Research & Model

Literature review, threat modeling, simulation, and feasibility analysis. Define constraints before writing code.

02

Prototype & Test

Rapid prototyping with hardware-in-the-loop testing. SITL/HITL validation before any field deployment.

03

Harden & Optimize

Performance profiling, adversarial testing, resource optimization for edge deployment. Make it survive.

04

Ship & Open Source

Production deployment, documentation, community release under permissive licenses. Build in the open.

Down to the metal

Custom firmware, real-time control, sensor integration. We write code that talks directly to silicon.

Flight Controllers & Autopilots

Custom firmware for STM32-based flight controllers. PID tuning, sensor fusion with Kalman filters, fail-safe logic for autonomous operations.

STM32F7 STM32H7 FreeRTOS PX4
📡

Sensor Payloads

Multi-sensor payload design: thermal cameras, LiDAR, gas sensors, SDR receivers. SPI/I2C/UART integration and real-time data pipelines.

FLIR Lepton VL53L5CX BME688 RTL-SDR
🔌

ESP32 & RP2040 Platforms

Networked sensor nodes, mesh telemetry, tactical IoT. Wi-Fi/BLE/LoRa connectivity for distributed sensing and C2 networks.

ESP32-S3 RP2040 LoRa SX1276 MQTT

FPGA & Custom Silicon

Hardware-accelerated signal processing, video encoding, and crypto. Xilinx and Lattice targets for latency-critical applications.

Xilinx Zynq Lattice iCE40 Verilog Yosys
🔒

Secure Boot & Hardware Security

Secure bootloaders, encrypted firmware updates, hardware root of trust. Anti-tamper design for field-deployed systems.

ATECC608 ARM TrustZone JTAG/SWD OTA updates
📊

Power & Battery Management

Custom BMS design, solar MPPT, ultra-low-power sleep modes. Maximizing endurance for long-duration autonomous missions.

BQ25703A LTC3119 MPPT LiPo 6S

Intelligence at the edge

AI that runs on 10 watts, survives GPS denial, and makes decisions in milliseconds. Not cloud-dependent. Not theoretical.

# Edge inference on Jetson Orin
from publicsphere.edge import TRTDetector

detector = TRTDetector(
  model="yolov8n-custom.engine",
  precision="FP16",
  target="jetson-orin-nano"
)

for frame in camera.stream():
  detections = detector.infer(frame)
  mavlink.send_target(detections)

# 45 FPS @ 8W power envelope
# 12ms end-to-end latency
  • On-Device Neural Inference

    Custom-trained models optimized for Jetson, Coral, and FPGA targets. No cloud dependency, no latency penalty.

  • Visual SLAM & Navigation

    GPS-denied navigation using stereo vision, IMU fusion, and learned depth estimation for indoor and contested environments.

  • Multi-Agent Coordination

    Distributed consensus algorithms for drone swarm operations. Mesh networking, task allocation, and collision avoidance.

  • LLM-Powered Mission Planning

    Natural language mission specification translated to executable flight plans. Integrating foundation models with real-time telemetry.

Building in public

We believe defense tech and critical infrastructure software should be auditable, forkable, and community-driven.

edge-detect

Lightweight object detection framework optimized for ARM/CUDA edge devices. TensorRT & ONNX Runtime backends.

Rust / Python MIT License

swarm-proto

Protocol implementation for multi-UAV coordination. Distributed consensus, mesh networking, and formation control.

C++ / ROS2 Apache 2.0

mavbridge

High-performance MAVLink bridge with WebSocket API. Real-time telemetry streaming for ground station applications.

Rust MIT License

sigint-tools

SDR signal processing toolkit. Spectrum analysis, protocol decoders, and RF fingerprinting for research and education.

Python / C GPL-3.0

vslam-lite

Minimal Visual SLAM implementation for resource-constrained platforms. Stereo + IMU fusion with loop closure.

C++ BSD-3

tactical-mesh

Ad-hoc mesh networking daemon for tactical communications. LoRa and Wi-Fi backends with automatic routing.

Rust / Zig Apache 2.0

Serious problems need serious engineering

Collaboration inquiries

We selectively partner on research programs, joint development, and defense innovation initiatives. If you're working on a hard problem in autonomy, edge AI, or ISR — reach out.