Signal Processing & Feature Engineering ML Engineer - KOS AI. Job in Palo Alto WestCoast-Jobs

Full Description

Job Description Job Description Company Overview KOS is revolutionizing diabetes care with the Argus Continuous Glucose Monitoring System - the world's first non-invasive, optical CGM wristband. Our breakthrough technology uses advanced photoplethysmography and machine learning to provide continuous glucose monitoring without needles, sensors, or patches. Position Overview We are seeking a highlyskilled Signal Processing & Feature Engineering ML Engineer to lead the development of our advanced multi-wavelength photoplethysmography (PPG) signal processing pipeline. You will be responsible for extracting glucose-specific information from complex optical biosignals, developing robust feature engineering algorithms, and ensuring clinical-grade accuracy across diverse populations and conditions. Key Responsibilities Advanced Signal Processing Pipeline Design and implement the 3-stage signal conditioning pipeline: - Stage 1: Pre-processing on AFE/MCU (dark-current cancellation, time-division multiplexing, anti-aliasing) - Stage 2: Physiological enhancement (motion artifact suppression, band-pass filtering, green-channel synthesis) - Stage 3: Windowing and feature extraction (500 physiological features across 5 domains) Multi-Wavelength PPG Optimization Develop algorithms for IR, Red, and Green wavelength processing Implement time-division multiplexing optimization with wavelength-specific parameters Create adaptive LED current and pulse width control algorithms Design ambient light cancellation and dark-current subtraction systems Motion Artifact Suppression Build adaptive least-mean-square (LMS) filters for accelerometer-based motion correction Develop real-time step noise attenuation algorithms (>18dB suppression in Implement 6-axis IMU integration for comprehensive motion compensation Create physiological constraint models for motion artifact detection Feature Engineering & Extraction Design comprehensive feature extraction across 5 domains: - Time Domain:

Peak-to-peak, rise/fall times, skewness, kurtosis, 32 statistical moments - Frequency Domain: PSD slope, spectral entropy, harmonics, band power ratios, wavelets - Morphology: Systolic-diastolic ratio, dicrotic notch index, 85 morphological indices - Non-linear: Sample entropy, Poincaré analysis, recurrence quantification, fractals - Glucose Dynamics: 200 glucose kinetics features with physiological constraints Multi-Dataset Harmonization Develop algorithms to integrate diverse data sources (Guilin, Mendeley, PPG-Dalia, Glucdict, proprietary datasets) Implement sampling rate normalization across 25-1000Hz acquisition rates Create channel synthesis for single-channel dataset augmentation Build equipment calibration compensation for variant hardware specifications Design signal quality assessment metrics for corrupted segment rejection Skin Tone Optimization Develop green-channel synthesis algorithms for melanin-induced SNR loss mitigation Implement adaptive linear mixture models: (t) = R(t) IR(t) Create calibration snippet collection and optimization algorithms Build Fitzpatrick skin type classification and adaptation systems Required Qualifications Education & Experience MS/PhD in Electrical Engineering, Biomedical Engineering, or Signal Processing 5 years of experience in biomedical signal processing 3 years of experience with photoplethysmography (PPG) or similar optical biosignals Proven track record in real-time signal processing algorithm development Technical Skills Signal Processing: Advanced DSP, filter design, spectral analysis, wavelet transforms Programming: MATLAB, Python, C/C++, embedded signal processing ML/Statistics: Feature engineering, time-series analysis, statistical signal processing Hardware Integration: AFE (Analog Front-End) programming, ADC optimization, sensor fusion Tools: MATLAB Signal Processing Toolbox, Python scipy/numpy, STM32 ecosystem Domain Expertise Deep understanding of cardiovascular physiology and PPG signal characteristics

Experience with motion artifact suppression and noise reduction techniques Knowledge of optical properties of biological tissues Familiarity with medical device signal processing requirements Preferred Qualifications Experience with continuous glucose monitoring or diabetes technology Background in optical sensing and photonics Knowledge of FDA/ISO standards for medical device signal processing Experience with multi-modal sensor fusion (PPG accelerometer gyroscope) Publications in biomedical signal processing or optical sensing Experience with real-time embedded signal processing constraints Technical Challenges You'll Solve Extracting glucose-specific signals from complex cardiovascular waveforms Achieving robust performance across diverse skin tones and physiological variations Real-time processing with 90ms latency constraints on embedded hardware Handling motion artifacts during daily activities and exercise Maintaining signal quality across temperature and humidity variations Scaling algorithms across 1000 hours of diverse training data Key Performance Metrics Accuracy: Contribute to overall system MARD Latency: Maintain 90ms processing time per 10-second window Robustness: >45% reduction in motion/ambient noise vs. baseline methods Memory Efficiency: Operate within 38kB RAM and 64kB flash constraints Signal Quality: Achieve >96% data capture rate in real-world conditions What We Offer Compensation & Benefits Competitive salary Equity package in a high-growth health tech company Comprehensive health insurance Growth & Impact Lead breakthrough research in non-invasive optical glucose sensing Direct impact on millions of people with diabetes worldwide Opportunity to publish research and present at top-tier conferences Collaboration with world-class biomedical engineers and clinicians Clear path to technical leadership and principal engineer roles Work Environment Access to state-of-the-art signal processing and optical measurement equipment Collaborative culture

with cross-functional medical device teams Regular interaction with clinical researchers and diabetes specialists Access to diverse clinical datasets and real-world validation studies Research & Development Opportunities Explore novel optical wavelengths and sensing modalities Develop next-generation motion compensation algorithms Investigate personalized signal processing adaptation techniques Contribute to patent applications and intellectual property development Collaborate on clinical studies and regulatory submissions Application Process Please submit: 1. Resume highlighting signal processing and biomedical experience 2. Cover letter describing your interest in optical glucose monitoring 3. Portfolio of signal processing projects (code samples, publications) 4. Any relevant patents, publications, or technical presentations Equal Opportunity KOS is an equal opportunity employer committed to diversity and inclusion. We welcome applications from all qualified candidates regardless of race, gender, age, religion, sexual orientation, or disability status. Ready to transform raw optical signals into life-changing glucose insights? Join us in pioneering the future of non-invasive diabetes monitoring. Company Description Founded five years ago in Palo Alto, we build everything in-house, including hardware, software, and machine learning systems. Company Description Founded five years ago in Palo Alto, we build everything in-house, including hardware, software, and machine learning systems.