MLEngineer

Skills for AI + ML Engineer

CORE APPLIED INTERVIEW DEPTH NEW

Minimum Viable Profile

6–8 months — must have shipped:

RAG system on real docs with Recall@5 + faithfulness eval · GitHub with README
7B model fine-tuned via QLoRA on specific task · before/after eval · HuggingFace
Agentic app (customer support / doc Q&A with tools) · deployed & accessible
SQL: solve medium LeetCode SQL in <15 min
FastAPI endpoint with pytest tests

global roles 12–15 months — additionally:

Can implement any Layer 3–5 paper from scratch
Designed AI system end-to-end with cost + latency estimates
Run a real A/B test with proper statistical methodology
Deployed vLLM/FastAPI+GPU serving under real traffic
CI/CD eval harness — promptfoo + RAGAS on every PR
Knows prompt injection vectors and can design mitigations
Opinions on model choice backed by benchmark evidence