Deep Agentic Reasoning Engineer (Lorenz Labs)

Analog Devices’ Lorenz Labs seeks a Deep Agentic Reasoning Engineer (open rank) to develop multimodal reasoning models for Edge AI applications. The role focuses on models that reason across time-series, audio, video and other data, contributing to the group’s research on agentic systems and physical intelligence. The engineer will collaborate with AI, NLP, audio and time-series experts to align research with business needs.

Responsibilities

• Build and evaluate deep learning models for multimodal classification and reasoning.
• Work across the research lifecycle: stakeholder discussions, data collection, annotation, preprocessing, model training and rigorous evaluation.
• Prototype in Python with PyTorch or TensorFlow; improve multimodal reasoning through mid-training and post-training techniques.
• Use knowledge distillation to transfer reasoning capabilities from larger models to smaller, efficient models.
• Document experiments and results in internal reports, blog posts, and conference or workshop publications.

Qualifications

• PhD in an AI-related field such as NLP, computer vision or speech processing, or a related discipline, and a strong publication record in top-tier research venues.
• Proficiency in building and evaluating transformer-based models; expertise in supervised fine-tuning and post-training for reasoning models; and experience with multimodal reasoning involving data such as text, images, audio and video.
• Familiarity with Chain-of-Thought reasoning, agentic reasoning systems, task orchestration, LangChain and LangGraph; knowledge of reinforcement learning approaches such as GRPO; and a strong understanding of LLM evaluation.

The position is listed in San Jose, California, and Wilmington, Massachusetts. Required travel is 10%, with a first-shift/day schedule. The expected new-hire wage range is $197,800 to $271,975, with eligibility for a discretionary performance-based bonus. Benefits include medical, vision and dental coverage, a 401(k), paid vacation, holidays and sick time. Applicants who are not U.S. citizens, permanent residents or specified protected individuals may need an export-licensing review for access to technical data.