Senior AI Engineer

About Repair Wise

Repair Wise is an AI-powered repair platform built for the EV era. The US automotive repair market exceeds $300B annually — and EV repair complexity is growing faster than the industry’s tooling can keep up. We partner with 90+ active shops (up from 18 less than a year ago), have processed over $4M in repair GMV, and are growing rapidly. Our small, high-ownership engineering team moves fast and builds things that matter in a market that’s changing by the month.

The automotive repair industry is ripe for AI disruption but deeply underserved by good tooling. Our agents handle real-world complexity: ambiguous diagnostic codes, warranty policy interpretation, technician-level reasoning about vehicle systems. We’re now moving beyond API-driven AI to training domain-specific models on our own repair data — a genuinely hard problem with real consequences. If you want to own the full AI stack at a company where your work directly shapes the product, this is the role.

We’re looking for a Senior AI Engineer to own and lead AI development across our platform. This is a hands‑on, high‑ownership role — you’ll architect, build, and ship our AI systems end‑to‑end, not just implement what others design. You’ll work directly with our Head of Engineering, have a seat at the table on product direction, and be the person who sets the technical bar for AI at Repair Wise.

Our AI work spans LLM orchestration, agentic workflows, and — increasingly — training domain‑specific models on our proprietary repair data. We’re not doing academic research. We’re building AI that helps shop technicians diagnose vehicles, service advisors write repair orders, and operators process warranty claims at scale.

Architect, build, and evolve the AI agents that power Repair Wise’s core workflows — from repair intake through resolution. You’ll own their design, reliability, and continuous improvement.

Train and fine‑tune domain‑specific models on our proprietary repair data for two high‑value use cases: automated repair plan creation and warranty claim writing. This includes owning the full pipeline — dataset curation, annotation, training, evaluation, and deployment.

We ingest and process vehicle telemetry and diagnostic data from OEM sources. You’ll help us build smarter, faster, and more accurate diagnostic systems as we expand our vehicle coverage.

Our analytics stack is built on DuckDB/Mother Duck. You’ll work with structured repair order data, labor codes, and shop performance metrics to train, evaluate, and improve model outputs.

Own evals, feedback loops, and guardrails for both LLM‑powered agents and fine‑tuned models — in a domain where accuracy has direct financial consequences for our shop partners.

(No specific stack details were provided.)

• 5–8 years of software engineering experience, with at least 3 years in applied AI/ML or LLM systems
• Proficiency in Python — our primary language for AI/ML work
• Hands‑on experience building and shipping LLM‑powered products — RAG pipelines, agents, tool‑calling, prompt engineering
• Experience with LLM APIs (Anthropic, OpenAI, or similar) and modern orchestration patterns
• Demonstrated experience fine‑tuning or training models using PyTorch and Hugging Face Transformers
• Experience owning dataset curation and annotation pipelines for model training
• Experience building evaluation and observability tooling for LLM systems — evals, feedback loops, and guardrails
• Ability to own systems end‑to‑end — from architecture decisions through deployment and monitoring
• Comfort working with structured data and SQL — our domain is rich tabular data, not just text
• Product instinct — you care whether the thing you built actually helps the user

• Experience with agentic frameworks (Mastra, Lang Graph, Llama Index, or similar)
• Familiarity with PEFT/LoRA or other parameter‑efficient fine‑tuning techniques
• Experience with experiment tracking tools (Weights & Biases, MLflow, or similar)
• Familiarity with vector databases (Pinecone, Weaviate, pgvector, or similar)
• Familiarity with…