Principal Engineer Tech Lead, Embodied AI & Off-Board Performance Evaluation

Motional•Boston, MA

34d•$200,000 - $275,000•Hybrid

About The Position

The Systems Readiness and Performance team is the crucial bridge between software development and real-world deployment. We are responsible for driving system design, for verifying and validating the autonomy stack, and for defining, measuring, and validating system performance targets. We work closely with stakeholders in autonomy, infrastructure, and operations to build the definitive safety case for the commercial launch of our fully driverless IONIQ 5 robotaxis in Las Vegas later this year. We are seeking a talented Principal Engineer to technically oversee the design, development, and deployment of a novel Embodied AI and Large Language Model (LLM)-based monitoring framework for off-board scenario understanding, intelligence evaluation, and root cause analysis. While this role will pioneer future off-board Vision-Language-Action (VLA) integrations for off-board analysis, it requires strong foundations as an Autonomous Vehicle Performance Metrics developer. You will lead the creation of a specialized, off-board evaluation layer that consumes AV drive logs and fleet data to automatically infer scenario safety and assess driving intelligence without running live on the vehicle. Your architectural designs will integrate with internal systems to form a complete pipeline that identifies safety incidents, enriches them with structured context, and conducts detailed root cause analysis to provide the Autonomy team with actionable direction. If you are an innovative individual contributor with a passion for overseeing the integration of complex, scalable data pipelines and state-of-the-art Embodied AI frameworks, we encourage you to apply.

Requirements

10+ years of professional experience in software engineering, applied AI/ML, or autonomous vehicle systems development.
Bachelor's degree in Computer Science, Engineering, Robotics, or a related field.
Proven experience working with Large Language Models (LLMs) and Vision-Language Models (VLMs) for reasoning, parsing, and scene description.
Experience with parameter-efficient fine-tuning and deploying open-weights models on internal infrastructure.
Familiarity with local and cloud vector databases, such as LanceDB, for housing output vector embeddings.
Experience with adversarial scenario generation and closed-loop simulation environments.
Strong background leveraging software to develop frameworks, libraries, and tools for calculating and aggregating AV performance metrics.
Strong analytical and problem-solving skills, particularly in the context of complex system performance evaluation.
Expert-level proficiency in Python and strong understanding of software development principles.

Nice To Haves

Experience working with autonomous vehicle sensor data, including its processing and integration.
Hands-on experience with data pipeline orchestration tools and distributed data processing frameworks.
Expertise managing cloud infrastructure on AWS or GCP for processing terabytes of data efficiently.
Familiarity with Ray and Ray clusters for scaling Python applications and AI/ML tasks.
Expertise with C++ programming for data frameworks.

Responsibilities

Technically oversee the architecture to identify, describe, and enrich events in historical vehicle logs using Multimodal LLMs.
Oversee the off-board ingestion and fusion of semantic scene descriptions, ego-centric kinematics, and internal autonomy telemetry to create a holistic diagnostic context for LLM inference.
Develop structured prompting templates utilizing Contextual Prompting (CP), Chain-of-Thought (CoT), and In-Context Learning (ICL) to evaluate scenarios.
Architect the integration of foundation models into the Metrics Engine (ME), designing efficient cascade filtering and log slice parallelization strategies to scale high-volume LLM inference across simulation and on-road drive logs while managing computational latency and costs.
Define, design, and implement key metrics to evaluate autonomous vehicle performance, such as lane change capability, oscillations, and braking.
Deploy and manage a Retrieval-Augmented Generation (RAG) vector database containing codified AV Driving Policies to ground off-board LLM evaluations in specific Operational Design Domains.
Serve as a technical escalation point and collaborate with Autonomy (Planner, Prediction, Perception) and Systems teams to deliver high-signal, enriched events.
Drive the transition toward Direct Vector-LLM Fusion utilizing emerging Physical AI ecosystems and open-weights Vision-Language-Action (VLA) models to process telemetry off-board without text-translation bottlenecks.