Edge AI ML Engineer

Bose Corporation, U.S.A•Framingham, MA

3d•Onsite

About The Position

As an Edge AI ML Engineer, you will develop, optimize, and deploy machine learning models for audio and multimodal intelligence on real devices. You will work at the boundary of ML modeling, DSP, and embedded systems, turning research concepts into efficient, robust, production-ready algorithms. You will collaborate closely with ML researchers, DSP experts, firmware engineers, and hardware teams to design algorithms that perform well not only in the lab, but also under real-world constraints such as latency, memory and power. This role is ideal for an ML engineer who enjoys model development, experimentation, and algorithm design, while also caring deeply about whether those models can run efficiently on edge hardware.

Requirements

Strong proficiency in C/C++ for embedded systems.
Strong experience developing and evaluating machine learning models, preferably for audio, speech, or other time-series sensor data.
Experience optimizing ML models for edge, embedded, or resource-constrained environments.
Proficiency in Python for model development, experimentation, evaluation, and tooling.

Nice To Haves

Master’s or Ph.D. in Computer Science, Electrical Engineering, Machine Learning, or related field.
Experience with ML compilers/frameworks such as MLIR, Glow, ExecuTorch.
Experience with real-time streaming inference pipelines.
Knowledge of acoustics and classical audio DSP.
Experience with on-device ML (TinyML, quantization, pruning).
Publication track record in ML, DSP, systems, or embedded AI.

Responsibilities

Identify opportunities for new DSP/ML algorithms by deeply understanding device constraints, sensor characteristics, and hardware capabilities (MCU, DSP, NPU).
Develop audio and multimodal ML models for embedded and edge AI applications.
Design, train, evaluate, and iterate on models for real-world sensing and interaction use cases.
Prototype novel approaches that push what’s possible in low‑latency, on‑device audio and multimodal processing.
Develop software for RTOS environments (e.g., FreeRTOS) and deploy models to device runtimes and hardware accelerators (DSP, NPU, MCU).
Convert trained ML models into efficient embedded implementations (C/C++, quantization, fixed‑point inference).
Optimize runtime performance: memory footprint, SRAM usage, latency, and power consumption.
Integrate ML inference into real‑time firmware pipelines.
Design end‑to‑end embedded AI systems (sensor → preprocessing → model → post‑processing).
Architect reusable embedded ML platform components usable across multiple hardware targets.
Profile and optimize performance on MCUs, DSP cores, NPUs, and custom accelerators.
Work with cross‑compilation toolchains, CMake‑based builds, and modular codebases.
Integrate with on‑device ML runtimes (e.g., TFLite Micro, ExecuTorch, custom interpreters).
Provide actionable feedback on model architecture for deployment efficiency and real‑time behavior.