Principal Applied Scientist - ADAS Engineering Department

Subaru•MI, MI

17h•$140,000 - $165,000•Onsite

About The Position

Functions as senior individual contributor establishing academically-credible applied research within Subaru Research & Development (R&D) and translating it into production‑relevant innovation for artificial intelligence (AI) modeling and autonomous driving. Serves as the department's most senior technical authority for AI modeling and autonomous driving, setting the scientific bar for rigor, evaluation, and adoption decisions across projects while partnering closely with Engineering to bring research into production. Acts as an external-facing research ambassador and collaboration lead with universities and partners; maintains an internal-first publishing posture while ensuring work meets credible academic standards and supports intellectual property (IP) creation.

Requirements

Ph.D. in machine learning, computer science, electrical engineering, robotics, or a closely-related field required
Up to 2 years required
At least one (1) year of relevant experience that reflects the following:
Demonstrated record of rigorous research (hypothesis → experiment → analysis → conclusion) with evidence of impact via peer-reviewed publications and/or granted patents; external publishing is not required as a primary output, but scientific credibility and defensible evidence are required.
Proven ability to translate research prototypes into production systems through deep collaboration with engineering teams, including evaluation, deployment considerations, and reliability tradeoffs.
Track record of influencing technical direction across multiple initiatives without direct authority, raising the technical bar through standards, reviews, and decision quality.
Deep expertise in modern artificial intelligence (AI) modeling relevant to autonomous driving (e.g., perception, multimodal/foundation models, robustness/uncertainty, or planning-related learning) with the ability to judge tradeoffs under real-world constraints.
Experimental Design and Evaluation Rigor: Proficiency in dataset construction, metric design, failure analysis, statistical reasoning, reproducibility, and clear technical writing.
Research-to-Product Execution: Ability to produce implementable guidance, baselines, acceptance criteria, and staged adoption plans suitable for engineering execution.
Technical Leadership and Communication: Ability to present complex technical issues to research, engineering, and leadership audiences and influence decisions with evidence.
Mentorship and Bar-Raising: Consistent ability to mentor and uplevel scientists/engineers via reviews, standards, best practices, and documentation.
Strong Python proficiency and practical familiarity with modern ML tooling sufficient to prototype, evaluate, and guide engineering implementation

Nice To Haves

Ph.D. with doctoral research aligned to autonomous driving and artificial intelligence (AI) modeling (e.g., perception/computer vision, multimodal learning, sensor fusion, robustness/uncertainty, planning-related learning, or safety evaluation) preferred
Experience in advanced driver-assistance systems (ADAS)/autonomous vehicles (AV), robotics, embedded machine learning (ML), or other high-reliability domains where safety, edge cases, and operational constraints are central Preferred
Comfort representing the organization in technical reviews and research collaborations with academia/industry partners while maintaining internal-first externalization Preferred

Responsibilities

Owns the applied research portfolio for AI modeling and autonomous driving, selecting problems that materially advance advanced driver-assistance systems (ADAS) safety and future autonomous capability and that can be translated into production-relevant innovation.
Defines and drives the scientific/technical strategy for complex, ambiguous problems by establishing success criteria, evaluation gates, decision points, and staged roadmaps that guide teams from hypothesis to production-ready evidence.
Formulates research hypotheses; designs experiments; defines datasets and evaluation metrics; and executes rigorous studies beyond routine development work, including failure analysis, robustness testing, and boundary-case characterization relevant to autonomous driving safety.
Translates external and internal research into production-viable solutions by adapting methods to real vehicle constraints (compute/latency, reliability, maintainability, data constraints) and Subaru engineering workflows, partnering with Engineering to ensure research becomes deployable systems rather than prototypes.
Creates clear “research-to-product” handoff packages including problem framing, baseline comparisons, experimental evidence, implementation guidance, operational considerations, and acceptance criteria that engineering teams can execute against.
Builds and sustains partnerships with academic labs by identifying the right collaborators, shaping joint research agendas, and maintaining a pipeline of emerging methods aligned to Subaru’s roadmap; represents Subaru in technical reviews and collaborations with an internal-first publication posture and selective external participation as approved.
Mentors and technically guides engineers and scientists by reviewing experimental design, raising scientific rigor, and developing internal best practices for research methods, evaluation, documentation, and reproducibility.
Drives IP creation by identifying patentable innovations and working with internal stakeholders to document novelty, defensibility, and application to Subaru programs.
Strengthens Subaru’s ability to attract PhD-level talent and collaborations by establishing a consistent research presence through high-quality internal artifacts, selective external engagements where approved, and active university partnerships.
Takes on urgent, technically-difficult cross-project initiatives when requested by leadership, decomposing ambiguous problems into tractable research plans and delivering high-confidence technical recommendations backed by evidence.