Applied AI Engineer
RemoteRemote, USProject-based, roughly 20–30 hours per week during an engagementProject rate set per engagement
Take LLM features from a demo that impresses to a system that holds up — retrieval that actually retrieves, evaluations that catch regressions, and costs that survive contact with real volume.
[ 01 ]The role
What this actually is
PropelAI runs retrieval and generation pipelines in production for clients and for its own products. The hard part is never the first output; it is the hundredth, on a document set nobody cleaned, at a price that still works.
This is a contract role attached to specific engagements. The work is concentrated, well-scoped and ends on a date.
What you’ll do
- Build and tune retrieval pipelines over messy client document sets
- Stand up evaluation harnesses with a real golden set, and defend a change with numbers
- Control cost and latency — model selection, caching, batching, knowing when a small model is enough
- Write the integration code that puts the pipeline behind a product surface
- Document what is fragile so the next person does not rediscover it
What we’re looking for
- Shipped at least one LLM-backed system that real users depended on
- Fluent Python; comfortable with vector stores, embeddings and reranking
- You measure before and after, and you know why a single anecdote is not evidence
- Able to work independently against a fixed scope and date
Nice to have, not required
- Local or self-hosted inference experience — Ollama, vLLM, quantized models on your own hardware
- Structured-extraction work on legal, regulatory or procurement documents
[ 02 ]Apply
Apply for Applied AI Engineer.
One form, one PDF, no account to create. Attach a resume, point us at something you built, and tell us what you want to work on. Every application goes straight to Mike.
PropelAI is an equal opportunity employer. We consider all qualified applicants without regard to any characteristic protected by federal, state, or local law. If you need an accommodation at any point in the process, say so in your application or email [email protected] and we will sort it out.