Role profile · Engineering

Applied AI / ML Engineer

A type of work, not one opening. Join it from your career profile and you’re matched to Applied AI / ML Engineer roles as they open. No match is guaranteed, and no company sees you until you say yes.

What strong looks like

Has a model-backed system in production and knows how it behaved after launch, including any degradation. Has an eval they trust and one they stopped trusting. Can say cost per run and latency without looking. Has made a retrieval, tool-use, or fine-tuning decision and can say what evidence would have changed it. Knows what the model cannot do and designs around it.

Core must-haves

• A model-backed system in production that they owned through at least one degradation
• Strong software engineering in Python plus one systems language; not notebooks only
• Evals: has designed one they trust and stopped trusting another; can say what a benchmark misses
• ML fundamentals: training, fine-tuning, retrieval, or inference at the level of making real trade-offs
• Communication: explains a model's behavior and limits to people who did not build it
• Cost and latency literacy: can say what a run costs and what they changed to move it

Across every profile: ownership in ambiguity; communication that survives a follow-up; ai-native practice; verifiable outcomes.

Good to have

• Research signal: a publication, a reproduction, or post-training work (routes to Research Engineer when it dominates)
• GPU, CUDA, or inference-serving depth
• Customer-facing deployment of AI systems
• Developer-facing work: SDKs, docs, community
• A vertical domain where the data is hard: legal, health, finance
• Data engineering: pipelines and synthetic data generation

Seniority

Open from recent graduates to 15 or more years. We read level from what you have owned, not from years or brand. What the evidence looks like at each stage:

New grad to 2 years. A model-backed project that ran for real users, an eval they wrote, and a clear account of what the model got wrong. Strong fundamentals matter more than tenure here; p25 is 2 years.

2 to 5 years. Owns a production AI system, has a degradation story, states cost and latency without looking.

5 to 10 years. Sets the architecture and the eval standards for a team, has reversed a retrieval or fine-tuning decision, directs others' model work.

10 to 15+ years. Sets how an organization applies models, chooses what not to build on a model, has done it across generations of tooling.