Apply on Employer Site

Analog Devices · 1 day ago

Principal Engineer – Time-Series & Sensor Reasoning Models (Lorenz Labs)

San Jose, CA

Full-time

Onsite

Lead/Staff

$171K/yr - $256K/yr

10+ years exp

Analog Devices, Inc. is a global semiconductor leader that bridges the physical and digital worlds. They are seeking a Principal Engineer to advance AI engineering at the intersection of sensing, signal intelligence, and large-scale temporal modeling, contributing to the development of time-series foundation models for context-aware reasoning across time.

DSPElectronicsLightingSemiconductor

No H1B

U.S. Citizen Only

Responsibilities

Lead R&D on time-series foundation models that integrate multi-sensor streams (e.g., audio, motion, environmental, and physiological)

Develop compact, recursive, and hybrid modeling approaches (e.g., Tiny Recursive Models, Liquid Neural Networks, State-Space Transformers) for efficient deployment on edge hardware

Advance research in sensor fusion, enabling cross-modal alignment between acoustic, inertial, and photonic domains

Explore audio reasoning models that interpret context and intent through dynamic acoustic and environmental cues

Create benchmarking pipelines for cross-domain time-series foundation models, covering representation robustness, interpretability, and hardware performance metrics

Apply alignment and fine-tuning methods such as LoRA, Q-LoRA, adapter-tuning, and contrastive alignment for multimodal sensor datasets

Investigate modern foundation alignment techniques, including DPO (Direct Preference Optimization) and RLAIF (Reinforcement Learning from AI Feedback) for physical and sensory reasoning tasks

Partner with ADI’s hardware, signal processing, and systems teams to co-design architectures for real-time, energy-efficient sensing applications

Publish and represent ADI at major ML and signal-processing venues (NeurIPS, ICLR, ICML, ICASSP, KDD), often in conjunction with leading AI industry partners

Mentor junior researchers and help shape Lorenz Labs’ strategy for foundation models that understand and reason about physical systems

Qualification

Time-series MLSignal processingFoundation modelsSensor modelingMultimodal reasoningRepresentation learningCausal inferencePythonPyTorchCollaborationMentoring

Required

Deep expertise in time-series ML, signal processing, and foundation models (Chronos, TimesFM, TimeGPT, etc.)

Strong background in sensor modeling and signal fusion (e.g., PPG, IMU, audio, photonics, or industrial sensors)

Experience in context-aware and multimodal reasoning—especially involving audio perception, biosignals, or environmental context

Proficiency in representation learning, causal inference, and motif discovery in high-dimensional temporal data

Familiarity with benchmarking, evaluation, and robustness testing of foundation and fine-tuned models

Proven hands-on expertise with modern alignment and fine-tuning strategies, including parameter-efficient fine-tuning, LoRA/Q-LoRA, and reward-based optimization methods (DPO, PPO, RLAIF)

Fluency in Python, PyTorch, and large-scale training pipelines using cloud or distributed systems (AWS, GCP, etc.)

Ability to collaborate across disciplines—ML, hardware, and embedded systems—and translate research into deployable physical intelligence systems