AI/ML developers

Typical projects

• LLM integration: RAG systems, chatbots, AI-powered features
• Custom model training for classification, NLP, or computer vision
• ML pipeline development (training, evaluation, deployment)
• Data preprocessing and feature engineering
• MLOps: model monitoring, A/B testing, drift detection

Team setup

AI/ML engineers are expensive and in high demand. For LLM integration (chatbots, RAG, AI features), a strong software engineer who has learned LLM patterns is often better value than a PhD ML researcher. For custom model training (recommendation systems, fraud detection, computer vision), you need genuine ML expertise. In both cases, pair ML engineers with backend engineers who can handle the production infrastructure. A common mistake is hiring ML engineers and having them build CRUD APIs — that's expensive misallocation.

What to look for

• For LLM work: experience with API integration (OpenAI/Anthropic), prompt engineering, RAG architecture, vector databases
• For traditional ML: strong fundamentals in statistics, feature engineering, model evaluation
• Can deploy models to production — not just train in Jupyter notebooks
• Understands evaluation: knows how to measure whether a model is actually working
• Software engineering skills alongside ML skills — clean code, testing, version control

Red flags

• Can't deploy anything — lives in Jupyter notebooks and hands off 'finished' models to engineering
• Chases state-of-the-art models when a simple heuristic or logistic regression would solve the problem
• No understanding of evaluation metrics beyond accuracy — precision, recall, F1, business-specific metrics
• For LLM work: treats prompt engineering as copy-paste from blog posts, no systematic approach
• Can't explain their model's behavior to non-technical stakeholders