Applied AI Engineer

About the role

MaintainX is a leading mobile-first work execution platform for industrial and frontline teams. More than 15,000 customers, including McDonalds, Michaels, Cintas, Redimix, and U.S. LBM, use MaintainX to cut unplanned downtime and run better operations, across 16 million managed assets and 92 million completed work orders.

In August 2026 MaintainX became part of Autodesk, joining Autodesk Operations Solutions https://adsknews.autodesk.com/en/views/inside-autodesk-operations-solutions/, the organization unifying Autodesk's operations platform alongside Tandem, FlexSim and Fusion Operations. Autodesk's strategy is to converge design, make and operate into one continuous lifecycle: design an asset, build it, run it, then feed what you learn running it back into the next design. Autodesk had design and make. Operate is the phase that tells you what actually happened, and it is ours.

Frontier models are getting commoditized. The operating data underneath them isn't — and that's what we own. 13.9M+ managed assets, 79.5M+ completed work orders, and 150,000+ technicians generating trustworthy operating data every week, at the point of work.

Document Intelligence is the horizontal engine that turns that raw material — any file a customer hands us — into structured, trustworthy maintenance knowledge, and into the AI-native entities and answers built on top of it.

The Role

You'll own the Generation half of Document Intelligence: turning multimodal primitives (keyframes, transcripts, OCR) into schema-valid entities like SOPs, and holding the line on quality so "fast" never becomes "fast and wrong."

- Design and iterate recipe prompts — system, few-shot, and context assembly — for each generation recipe

- Define per-entity target schemas and domain validators (procedure step-type rules, field caps) that generated output has to pass

- Build generation-quality eval datasets and rubrics, offline and online, running on LLMX's eval pipeline, and close the loop when quality regresses

- Assemble multimodal context windows from keyframes, transcript, and OCR so each model call has exactly what it needs

- Choose the model and token budget per recipe based on quality, cost, and latency tradeoffs

You'll ride on LLMX for model access and on Attachments for ingestion, and hand off validated entities to the domains that own them. You'll report to our Engineering Lead and work closely with the Processing side of Document Intelligence.

Minimum Requirements:

- Strong applied GenAI craft - prompt engineering, structured output / tool-use, RAG and retrieval-context patterns

- Real eval discipline: you've built datasets and rubrics, measured factuality/relevance/quality, and closed the loop on regressions yourself

- Shipped LLM features into production services, not notebooks, and can connect model performance to product impact

- Comfort working with multimodal inputs - video, PDF, audio, image - converted to text or structured output