Curiosity is
the starting point.

I’m Jonathan, an AI engineer focused on developing intelligent systems that perceive, understand, and interact with the physical world.

My background spans computer vision, 3D perception, reinforcement learning, simulation, and high-performance edge AI. I’ve led the development of end-to-end ML systems—from synthetic data generation and distributed training to inference optimization and deployment on real-world platforms.

I’m particularly interested in world models, embodied intelligence, and learning-based autonomy, with a long-term focus on advancing general-purpose AI and its applications in robotics, scientific discovery, and human well-being.

LinkedIn GitHub Email

Exploring what’s possible.
Engineering what’s next.

Explore the work↘

Huntsville, Alabama ✳Fig. 001 — Emergence

02 / RésuméBackground

The story
so far.

From mathematical foundations to intelligent systems in the physical world.

Experience

January 2024 – Present

AI Engineer

Cahaba Federal Solutions

Huntsville, Alabama

Lead AI engineer on the U.S. Army SMDC’s $15M Autonomous Aimpoint Selection and Tracking (A2ST) program, developing a system-agnostic edge AI platform for real-time autonomous perception and control.
Architected an end-to-end ML pipeline spanning Blender-based synthetic data generation, distributed PyTorch training with Ray and Kubernetes, and model compilation and optimization for GPU and edge accelerators.
Profiled training and inference workloads with PyTorch Profiler, NVIDIA Nsight, and TensorRT to identify compute and memory bottlenecks and optimize performance on deployment hardware.
Collaborated with hardware, optical, and systems engineers to integrate the platform into four fielded high-energy laser systems.
Prototyped 3D shape estimation, 6-DoF pose estimation, and differentiable-rendering-based pose refinement for electro-optical imagery.

May 2021 – January 2024

Machine Learning Engineer

Radiance Technologies

Huntsville, Alabama

Developed real-time neural object detection and tracking models for the U.S. Army’s Affordable Laser Weapon Module (ALWM), outperforming traditional tracking algorithms across two successful field demonstrations.
Built and optimized a multithreaded C++ inference runtime; combined architecture design, structured pruning, quantization, and runtime optimization to achieve 1,000+ FPS for FPGA deployment.
Contributed to an ONNX-to-AMD/Xilinx Versal AI Engine compiler written in C++, implementing distributed Conv2D and ReLU kernels across FP32, FP16, and INT8.
Designed a modular OpenGL simulation platform used across three funded programs to generate synthetic imagery for neural network training and algorithm evaluation.