AI Team
A dream team setup that unleashes talent to build AI products
Before LLMs
Before LLM APIs, AI teams focused on specialists for model development from scratch.
AI Product Manager: Defines product vision and metrics. Aligns AI with business goals
Data Scientist: Designs AI/ML models. Runs experiments and analyses data. Builds and tunes models. Maintains model performance
ML Engineer: Bridges data science and production. Scales models in production. Integrates models into codebases
Data Engineer: Sets up data pipelines. Ensures clean data for training. Provides scalable infrastructure.Lays the foundation for data-driven AI.
Software Engineer (normally sits outside the model team): Integrates AI with frontend and backend. Connects data science to user experience. Delivers speed and functionality.
With LLMs
With LLM APIs, teams evolve towards integration-focused workflows while maintaining core expertise.
AI Product Manager: Defines vision and metrics. Aligns capabilities with business goals
Software Engineer: Builds UI and backend logic. Integrates AI features. Handles data flow between UI, backend, and APIs. Maintains product performance.
ML Engineer: Evaluates quality of LLM outputs. Ensures alignment with goals. Identifies bugs and performance issues. Tests for accuracy and bias
AI Engineer: Builds full-stack AI applications. Uses LLMs to solve real problems. Applies LLMs to products. Measures AI performance. Performs prompt engineering.
Data Scientist (evolved role): Adapts expertise to LLM evaluation and optimisation. Analyses LLM performance patterns. Designs experiments for prompt effectiveness. Maintains statistical rigour in AI assessments.
Advanced Teams
For fine-tuning and advanced features, specialist roles expand their responsibilities.
Data Scientist (expanded scope)
Fine-tunes models with company data
Evaluates LLM performance
Runs experiments with different models
Identifies and fixes biases
Bridges traditional ML with LLM capabilities
Data Engineer (enhanced role)
Manages pipelines for fine-tuning
Ensures data quality
Maintains data lakes and warehouses
Supports real-time data via prompts
Adapts infrastructure for LLM workflows
ML Engineer (evolved responsibilities)
Integrates LLM API calls into production
Ensures scalable API integration
Sets up MLOps for LLM integration
Monitors API performance
Combines traditional MLOps with LLM operations
AI Engineer
The current job market (2025) shows AI engineer positions require hybrid skill sets, combining traditional ML expertise with LLM integration capabilities. Organisations value both foundational AI/ML engineering skills and modern LLM application abilities.
Scaling Infra
Essential capabilities for scaling AI.
LLM Accessibility
Provides easy LLM access
Standardises integration
Reduces adoption barriers
Observability Platform
Monitors AI performance
Tracks usage and costs
Identifies optimisation opportunities
Evaluation Support
Measures AI effectiveness
Provides testing frameworks
Ensures quality standards
Feedback Loop
Captures user interactions
Enables continuous improvement
Informs feature development
Experimentation Sandbox
Provides a safe testing environment
Enables rapid prototyping
Allows risk-free exploration
Recipes and Demos
Offers implementation examples
Accelerates development
Models best practices
Transitioning
💡 What types of leaders do you need to transition from a traditional team to an AI team?
Operator — focused on maintenance and functioning, through stability and security.
Guide — focused on humanistic growth and supporting others, through affiliation and bonding.
Achiever — focused on victory and success, through sheer drive and achievement.
Catalyst — focused on transformation and innovation, through inspiration and engagement.
💡 Leaders need to cover all four areas, but which ones should they focus on?
Tech Lead EM: tech-led, people + delivery
Team Lead EM: people-led, tech + delivery
Delivery EM: delivery-led, tech + people
Product EM: product-led, tech + delivery. (I think this can be a technical product manager)
Lead of Leads EM: although they are team and delivery focused, you need to be one of 1-4 before you can be a lead of leads.
Last updated