Sr. Embedded Machine Learning Engineer

Allen Control Systems•Austin, TX

About The Position

Allen Control Systems (ACS) is seeking a Senior Embedded Machine Learning Engineer to manage the complete lifecycle of deploying trained machine learning models, including supporting code, onto resource-limited edge hardware. This role bridges machine learning, embedded systems, and hardware engineering. The primary responsibilities involve optimizing models for strict constraints in latency, memory, power, and thermal budget, and developing the C++ code for on-device model execution and pre/post-processing. This is a highly cross-functional role requiring collaboration with CVML teams, embedded and firmware engineers, and product teams to ensure models are accurate, fast, small, and reliable in real-world applications.

Requirements

A Bachelor's or Master's Degree in Computer Science, Electrical Engineering, Computer Engineering or a related field, or equivalent practical experience.
10 or more years of professional software or systems engineering experience.
At least 2 years focused on deploying ML models to embedded or edge devices.
Very strong proficiency in C/C++ (most important) or Python.
Proficiency with CUDA.
Hands-on experience with PyTorch.
Hands-on experience with at least one edge runtime or inference format (TensorFlow Lite, ONNX Runtime, TensorRT, or similar).
Practical experience with model optimization techniques such as quantization (post-training and quantization-aware), pruning, or distillation.
Demonstrated ability to profile and optimize for latency, memory, and power on constrained hardware.
Working knowledge of embedded or edge platforms (e.g., NVIDIA Jetson, Google Coral, Qualcomm, ARM Cortex, or comparable NPUs and SoCs).
Working knowledge of Linux or an RTOS.
Solid grasp of computer architecture concepts relevant to inference, including memory hierarchy, fixed-point arithmetic, and accelerator offload.
Domain experience in computer vision or sensor processing on device.

Responsibilities

Model optimization: Apply quantization, pruning, knowledge distillation, operator fusion, and graph optimization to reduce model size and inference cost while maintaining accuracy.
Model conversion and deployment: Convert trained models into formats suitable for edge runtimes using ONNX and TensorRT, and deploy them to target hardware.
Hardware bring-up and benchmarking: Profile inference on accelerators (GPUs, NPUs, DSPs, TPUs, FPGAs), measure latency, throughput, memory footprint, and power, and drive necessary changes to meet targets.
C++ application integration: Design, write, and maintain C++ code for on-device inference, including application and library code, pre/post-processing pipelines, data and memory management, threading, and interfaces to the embedded system. Ensure the combined stack meets real-time constraints, memory budgets, and reliability on the target platform, using Python for tooling and validation where appropriate.
Accuracy and quality validation: Build test harnesses to verify on-device accuracy against reference results and detect regressions from optimization or quantization.
Model update pipeline: Contribute to tooling and processes for packaging, versioning, and delivering model updates to deployed devices, including over-the-air updates.
Cross-functional collaboration: Work with research, firmware, and product teams to set realistic performance targets and provide feedback on hardware constraints for model design.
Technical leadership: Establish best practices for edge deployment, conduct design and code reviews, and mentor other engineers on embedded ML techniques.
Development and optimization of computer vision algorithms for autonomous drone detection, tracking, and classification in real-time.
Design and implement machine learning models for resource-constrained environments with high accuracy and reliability.
Collaborate with electrical engineers to integrate computer vision systems into the turret's hardware architecture.
Conduct extensive testing and validation of computer vision algorithms in various scenarios to ensure robustness and performance.
Contribute to hardening the prototype turret into a military-grade system and assist in developing variants.