Researcher- Applied AI and Machine Learning

Cohu•San Diego, CA

12h•$125,000 - $175,000

About The Position

Cohu, Inc. is establishing an AI group within Semiconductor Testing to address and implement solutions for complex, real-world challenges. This role involves tackling problems that lack immediate, clear-cut answers, developing tools to assess solution effectiveness, and assisting field and applications engineers in significantly reducing their workflow times. As a newly created research position within the AI Group, this role focuses on solving ambiguous, real-world AI problems specific to semiconductor testing. The position requires end-to-end research and delivery on these complex AI challenges in semiconductor test. This includes transforming unstructured and varied inputs—such as vague or customer-specific test terminology, undocumented protocols, and extensive technical documentation—into functional models, retrieval systems, and evaluation methods suitable for practical application. Given the absence of a predefined approach, a significant portion of this role will be dedicated to defining the problem and establishing success metrics, alongside the problem-solving itself. Examples of the work involved include standardizing inconsistent, customer-specific test names and undocumented protocols into an internal structured test library; automating undocumented manual test procedures; and creating a RAG-based retrieval engine to provide engineers with cited, source-linked answers from comprehensive tester documentation, thereby eliminating the need for manual searching.

Requirements

Advanced degree in Computer Science with specialization in ML/AI
Advanced degree in physics, electrical engineering, signal processing, or a related scientific domain preferred.
3–5 years of applied ML/research experience, or equivalent depth from an advanced degree in a quantitative field.
In-depth understanding of the scientific method: ability to design experiments, form hypotheses, and both ask and answer ambiguous questions independently.
Applied experience with modern ML/AI techniques: retrieval-augmented generation and knowledge-graph retrieval, model fine-tuning, and agent frameworks.
Experience designing evaluation methods and benchmarks for AI/ML systems, with the statistical rigor to know when a result is real versus noise.
Strong programming fundamentals and Python fluency, sufficient to build and iterate.
Working familiarity of AI coding agents (Claude Code, Cursor, Codex, Antigravity, or similar) by default, not because a policy directs you to.
Demonstrated ability to scope and solve ambiguous technical problems independently.
Excellent written and verbal communication; must be able to explain research findings and tradeoffs to cross-functional partners who don’t share your technical background.
Semiconductor industry familiarity, particularly in semiconductor testing, ATE, or EDA tools preferred.
Understanding of software engineering fundamentals and best practices.
Evidence of self-led learning into new technical areas outside your formal training.
Exposure to GPU/server hardware and at least one LLM inference or fine-tuning runtime (e.g., vLLM, Hugging Face Trainer/PEFT).
Knowledge of C++ for performance-critical or hardware-adjacent work preferred.

Responsibilities

Collaborate with the engineering team and domain experts to evaluate and scope ambiguous problems, turning a vague pain point into a concrete, measurable target.
Learn existing programs, protocols, and domain documentation deeply enough to separate the real constraint from an assumed one.
Design experiments and establish success criteria and evaluation benchmarks aligned with business goals, with the statistical rigor to defend a result.
Develop and fine-tune models, algorithms, and retrieval/knowledge-graph methods to evaluate and improve proposed solutions.
Design and build agent systems and the RAG/knowledge-retrieval pipelines that connect AI reasoning to domain-specific toolsets.
Work with the engineering team to deploy rapid solutions into production, not just research prototypes.
Stay ahead of emerging ML/AI research and tooling so the group isn’t surprised by what competitors or customers adopt next, and bring back what’s actually usable, not just novel.
Explore emerging platform capabilities as the system matures, proposing new project directions rather than only executing assigned ones.