Junior AI Engineer

Join to apply for the Junior AI Engineer role at Onto Innovation
.

Onto Innovation is a leader in process control, combining global scale with an expanded portfolio of leading‑edge technologies that span the semiconductor value chain. Our breadth of offerings helps customers solve difficult yield, performance, quality, and reliability issues, optimizing their critical path of progress with smarter, faster, and more efficient solutions.

Junior AI Engineer

Team: Partner with our AI Lead Engineer and collaborate with field/service engineers supporting inspection & metrology tools across fabs.

Goal: Build practical AI helpers that speed tasks from recipe setup and troubleshooting to fleet management analytics and expert guidance from internal knowledge.

• Prototype AI assistants & agents for field workflows: guided recipe setup, log triage, playbook lookups, parts/alarms advice, and fleet‑wide health checks.
• Build retrieval systems (RAG): ingest manuals, specs, ticket notes, recipes, logs, and best‑practice docs; design chunking, embeddings, and indexing; tune prompts and retrieval for accuracy/latency.
• Connect AI to our tools and data: stand up MCP servers (Model Context Protocol) and other connectors to safely expose internal systems (document stores, MES, issue trackers, telemetry APIs) to LLMs.
• Fine‑tune or adapt models (e.g., LoRA/QLoRA) for domain terms, error codes, and tool‑specific intents when retrieval alone isn’t enough.
• Evaluate and harden: set up offline & online evals for groundedness/relevance; add guardrails, observability, and traceability; write runbooks.
• Ship small apps: package prototypes behind simple APIs or lightweight UIs that field engineers can use (web chat, Slack/Teams bots, or CLI).
• Data plumbing: parse messy PDFs/images/CSVs; normalize schemas for recipes, events, alarms, SPC/trace data.
• Computer Vision – understanding, defect detection, segmentation, or SEM/optical imaging.
• Work like an engineer: write readable Python/Type Script, tests, and docs; use Git; participate in code reviews; iterate fast with the AI lead and domain SMEs.

• BS in CS/EE/CE/ME (or equivalent experience).
• Python proficiency (data wrangling, APIs, packaging); comfort on Linux and with Git.
• Built at least one LLM app using a framework such as Lang Chain, Llama Index, or Semantic Kernel.
• Hands‑on with vector search (e.g., FAISS/Weaviate/Milvus) and embeddings; understands chunking, metadata, and hybrid search basics.
• Familiarity with RAG and prompt engineering; can measure quality (groundedness/relevance) and reduce hallucinations.
• Basic backend skills (REST/JSON, auth, environment secrets); experience containerizing with Docker.
• Comfortable reading technical manuals/logs and collaborating with non‑software teammates.

• Worked with agent frameworks (Lang Graph, Auto Gen, CrewAI) or implemented tool‑calling/plan‑execute loops.
• Built or configured MCP servers to connect LLMs to internal data/tools.
• Experience parsing complex docs (e.g., Unstructured, GROBID) and handling images/figures from manuals.
• Exposure to semiconductor equipment or factory systems (SECS/GEM, EDA/Interface A, MES, SPC); familiarity with KLA/AMAT/TEL/ASML tool ecosystems.
• Time‑series and log analysis (Pandas, SQL; Timescale
  
  DB/Influx
  
  DB), wafer map/vision background, or simple CV.
• Model adaptation experience (LoRA/QLoRA, PEFT) and experiment tracking (MLflow/W&B).
• LLM observability/evals (Ragas, Tru Lens, Lang Smith), basic security/PII handling, and role‑based access.
• Cloud familiarity (AWS/Azure/GCP) and lightweight front‑ends (React/Next.js) for internal tools.
• Prior work on fleet‑level dashboards/analytics or recipe/parameter management.

• Ship a search+chat knowledge assistant over our internal docs with clear eval dashboards for faithfulness/relevance.
• Stand up at least one MCP connector to an internal source (e.g., SharePoint/Confluence or log store) and demo safe tool calls.
• Deliver a focused POC: e.g., an agent that reads recent alarms & logs to suggest next steps, or a fleet health summary with links to playbooks.
• Document everything…