Senior Technical Architect

Job Summary

We are building a team to take research-grade audio algorithms and turn them into shippable, production-quality embedded software for ultra-low-power, AI-enabled microcontroller (MCU) platforms. An algorithm research group develops and validates audio algorithms (e.g., noise cancellation, wake-word detection, audio pipeline and ML models) as reference implementations — today largely delivered as MATLAB/prototype models that require conversion to embedded C. This role owns that productization: converting, optimizing, and adapting these algorithms into robust embedded C that runs across target hardware platforms and is tuned for specific end markets.

Key Responsibilities

• Convert audio algorithm reference implementations (e.g., MATLAB/prototype models) into optimized, production-quality embedded C — this is a core, ongoing responsibility of the role, not a one-time task.
• Productize algorithms such as noise cancellation, wake-word detection/adaptation, and combined audio-pipeline-plus-ML models for real-time, always-on operation.
• Perform platform-specific adaptation: partition and tune algorithms across a heterogeneous compute architecture where heavy signal-processing computation is offloaded to a DSP and ML inference runs on a dedicated NPU/AI accelerator — optimizing for low power and cost efficiency.
• Perform market-specific adaptation: tune and validate algorithms (e.g., wake-word variants, acoustic environments, language/locale requirements) for different regional or customer-market requirements.
• Design, implement, and optimize embedded audio signal processing pipelines (filtering, feature extraction, beamforming, noise/echo suppression) in C/C++ for MCU-class hardware.
• Port and optimize machine learning inference (quantized neural network models) onto microcontroller NPUs/DSP extensions, balancing accuracy, latency, memory footprint, and power consumption.
• Work with audio codecs, PDM/I2S digital microphone interfaces, and analog front-end configuration for multi-microphone arrays.
• Profile and optimize real-time audio pipelines against strict RAM/flash and power budgets typical of battery-operated or always-on devices.
• Collaborate closely with the upstream algorithm research team to clarify algorithm intent and validate that productized C implementations preserve reference behavior and audio quality.
• Write unit/integration tests, support bring-up on evaluation boards, and debug using JTAG/SWD and logic analyzers.
• Contribute to technical documentation, design reviews, and code reviews; help establish reusable productization processes and tooling as the team scales.

Skill Requirements

• Strong proficiency in embedded C/C++ and real-time, resource-constrained software development.
• Hands-on experience converting algorithm prototypes (e.g., MATLAB, Simulink, or Python reference models) into optimized, production embedded C — this specific skill is a must-have, not just general C/C++ ability.
• Practical understanding of digital audio signal processing: FFT/filtering, feature extraction (e.g., MFCC), noise reduction, acoustic echo cancellation, wake-word/keyword-spotting pipelines.
• Hands-on experience with Arm Cortex-M series microcontrollers (or similar embedded cores), DSP extensions, and associated toolchains (GCC/Arm Compiler, debuggers).
• Experience integrating or optimizing on-device (TinyML) inference — familiarity with frameworks such as TensorFlow Lite Micro, CMSIS-NN, or equivalent embedded ML runtimes, and with NPU/AI-accelerator offload.
• Working knowledge of RTOS concepts (task scheduling, low-power modes, interrupt handling) — e.g., FreeRTOS or similar.
• Experience with I2S/PDM audio interfaces and low-power always-on sensing architectures.
• Comfortable with hardware bring-up, oscilloscope/logic analyzer debugging, and reading datasheets/reference manuals.
• Version control (Git) and familiarity with CI-based embedded build/test pipelines.

Other Requirements

• MATLAB/Simulink literacy sufficient to read and faithfully translate reference algorithm models.
• Experience with model quantization/compression for edge inference.
• Exposure to multi-microphone array processing or beamforming.
• Familiarity with power-profiling tools and ultra-low-power design techniques.
• Prior experience in consumer IoT, wearables, smart home, or voice-assistant product development, especially taking a product from one market/region to another.
• Python scripting for tooling, data analysis, or model evaluation.
• MATLAB/Simulink literacy sufficient to read and faithfully translate reference algorithm models.
• Experience with model quantization/compression for edge inference.
• Exposure to multi-microphone array processing or beamforming.
• Familiarity with power-profiling tools and ultra-low-power design techniques.
• Prior experience in consumer IoT, wearables, smart home, or voice-assistant product development, especially taking a product from one market/region to another.
• Python scripting for tooling, data analysis, or model evaluation.

• Bachelor’s or Master’s degree in Electronics, Computer Engineering, Computer Science, or related field.