Build · Ship · Prove
Choose the work you want to do, turn skills into three inspectable projects, and decide where cloud or certification belongs after you have built the evidence.
The operating idea
All three paths share Python, Git, data literacy, APIs, evaluation, security, and documentation. Your primary path determines which evidence you build first and what you deliberately postpone.
Three primary paths
Choose this path when you want to train, adapt, evaluate, or optimize models and enjoy investigating why results change.
A complete experiment with a baseline, held-out evaluation, error analysis, and reproducibility record.
Review this path’s current milestonesChoose this path when you want to build useful software around model capabilities and care about users, interfaces, evaluation, and delivery.
An evaluated AI application with documented boundaries, failure behavior, authorization, latency, and cost observations.
Review this path’s current milestonesChoose this path when your strongest interest is making AI workflows reliable, repeatable, deployable, observable, and recoverable.
A bounded service another engineer can build, run, inspect, recover, and evaluate without relying on your memory.
Review this path’s current milestonesPortfolio sequence
Each project adds a different engineering responsibility. The templates keep decisions, evaluation, failures, and limitations visible.
Open project templatesTurn a defined decision into a measurable prediction problem, establish a baseline, evaluate honestly, and explain what the evidence does not establish.
Build an application with a user problem, information boundary, evaluation set, error policy, and an explicit response when the model is uncertain.
Make deployment, observability, reproducibility, security, and failure recovery explicit enough for another engineer to reason about the service.
Reader resources
These tools are optional. They help you save a decision, work with a template, or verify information that may change after publication.