Build your portfolio
Guided projects
Each roadmap stage ends in a project that proves you can do the work. Follow the steps, check yourself against the rubric, and share the link.
26 projects
- Advancedabout 24 hours6 stepsThe stage-3 agent, deployedDeploy your operations agent to a cloud with CI/CD, tracing, a cost dashboard and a gateway with a fallback model, and show it meets a stated latency budget under load.Forward Deployed Engineer (FDE): Production and deployment
- Intermediateabout 14 hours6 stepsAn AI feature briefWrite the product brief for one AI feature in an Indian consumer app, covering the problem, design, eval plan, metrics, pricing, risks and launch plan, with a prototype and 20 eval cases.AI Product Manager: Pricing, launch and growth
- Advancedabout 30 hours6 stepsAn AI platform for one teamStand up a gateway, an open-weight model and a hosted model behind it, with tracing, cost dashboards, CI/CD for prompts and an incident runbook, all built from infrastructure as code.MLOps Engineer: Governance
- Advancedabout 10 hours6 stepsA written design for an AI product, with interview practiceWrite a two-to-four page design for an AI product of your choice, covering users, metrics, components, failure modes, evaluation and cost, then answer ten interview questions about it out loud to a friend.AI Engineer: System design and interviews · Generative AI & LLM Engineer: AI system designer
- Intermediateabout 15 hours6 stepsAn AI risk assessment for a real AI featureAssess one real or realistic AI feature end to end, covering what it does, who could be harmed, which laws and policies apply, the tests that show it is safe enough, and the controls and owners, then present a launch recommendation.AI Governance & Compliance Specialist: Assurance and audit
- Starterabout 6 hours6 stepsA script that calls a public API and is pushed to GitHubWrite a Python script that calls a free public API, saves the results to a CSV, summarises them with pandas and lives in a GitHub repository with a clear history of small commits.AI Engineer: Python and data foundations · Generative AI & LLM Engineer: Data and tools
- Starterabout 3 hours6 stepsThree checked prompts for a real work taskPick one task from your work or studies, write three prompts that get an AI assistant to help with it, and check every answer against a trusted source. You finish with a short write-up a manager could read.Generative AI & LLM Engineer: AI newcomer
- Starterabout 6 hours6 stepsA script that cleans a messy CSVWrite a Python script that reads a messy CSV file, fixes common problems such as blank rows, stray spaces, mixed date formats and duplicates, writes a clean file and prints a short report of what it changed.Generative AI & LLM Engineer: Python starter
- Advancedabout 14 hours6 stepsA deployed app with monitoring and a cost reportDeploy one of your LLM apps so others can use it, with secrets kept safe, logging of latency, errors and cost, caching where it helps, alerts, and a one-page cost report for a month of expected use.AI Engineer: Production · Generative AI & LLM Engineer: Production engineer
- Intermediateabout 14 hours6 stepsA document question-answering APIBuild a FastAPI service that takes a PDF, extracts its text and answers questions about it with a hosted model, packaged with Docker so a stranger can start it with one command.Forward Deployed Engineer (FDE): Engineering foundations
- Advancedabout 24 hours6 stepsAn enterprise RAG assistant with an eval reportBuild a RAG assistant over a few hundred pages of public policy documents, with citations and refusals, and measure it with a golden question set and an eval report comparing two retrieval set-ups on accuracy, cost and speed.Forward Deployed Engineer (FDE): LLM engineering and RAG
- Advancedabout 12 hours6 stepsAn eval harness that shows a measured improvementBuild an evaluation harness for an LLM app you already have: a test set, automatic checks, a calibrated model judge and a regression gate in CI, then use it to show one change that measurably improves the app.Generative AI & LLM Engineer: Quality engineer
- Advancedabout 40 hours6 stepsAn end-to-end client capstonePick a real public dataset and an imagined client. Write the discovery notes, scope, architecture and KPI, build the system, and present the results as you would to the client's leadership.Forward Deployed Engineer (FDE): FDE craft and capstone
- Intermediateabout 8 hours6 stepsA small app that uses an LLM APIBuild a small command-line or web app that uses an LLM API for one useful job, returns structured output your code can check, handles errors and retries, and reports what each request costs.Generative AI & LLM Engineer: LLM builder
- Advancedabout 30 hours6 stepsA fraud model in productionTrain a fraud model on a public transactions dataset, choose a cost-weighted metric, serve it behind an API, monitor it for drift and document it in a model card.Machine Learning Engineer: Career
- Advancedabout 16 hours8 stepsImprove a coding agent through its harnessTake a small coding agent, build a fixed set of 20 tasks with automatic checks, measure a baseline, then change only the harness, one part at a time, and report what each change did to pass rate, cost and time.Agentic AI Engineer: Reliability
- Intermediateabout 14 hours6 stepsAn invoice-to-ledger automationBuild a workflow that reads invoices from an inbox, extracts the fields with an AI step, asks a person to approve anything unusual and writes the result to a spreadsheet, then measure its error rate and ROI.AI Automation Specialist: Delivering automation projects
- Intermediateabout 12 hours6 stepsA multi-step agent with logsBuild an agent that uses at least three tools to complete a multi-step task, logs every step it takes, stops safely when something goes wrong, and is tested on ten tasks including two it should refuse.Generative AI & LLM Engineer: Agent builder
- Advancedabout 22 hours6 stepsAn operations agent with approval gatesBuild an agent that triages support tickets, drafts replies and refund actions, and waits for a person to approve anything that changes a customer record, with traces, a budget and a kill switch you can show working.Forward Deployed Engineer (FDE): Agentic AI
- Starterabout 3 hours5 stepsA personal AI workflowPick one task you do every week, use an AI assistant to do it faster, check the result, and write down the steps so a colleague could repeat them.AI Governance & Compliance Specialist: How AI works and fails
- Intermediateabout 16 hours7 stepsA phishing email detector with honest evaluationTrain a small machine learning classifier that flags phishing emails, measure it honestly with precision, recall and false positives at a realistic base rate, add an LLM second opinion for borderline emails that treats the email as untrusted data, and write up what it catches and what it misses.AI Security Engineer: Securing LLM apps
- Intermediateabout 12 hours6 stepsA RAG app over 20 documents with a test setBuild a question-answering app over at least 20 public documents: split them into chunks, embed and search them, answer with citations, refuse when the documents are silent, and measure it with a 20-question test set.Generative AI & LLM Engineer: RAG builder
- Advancedabout 20 hours7 stepsRed-team your own LLM appRun an authorised red-team exercise on your own RAG chatbot. Threat model it, attack it with manual tests and an open-source scanner such as promptfoo or garak, write a findings report, fix the worst issues, and turn every successful attack into a regression suite that runs in CI.AI Security Engineer: Red-team and test AI
- Advancedabout 22 hours6 stepsA research agent with approvalsBuild an agent that researches a company from public sources, drafts a briefing and asks for approval before sending it, with traces, a budget, a kill switch and a trajectory eval over 20 tasks.Agentic AI Engineer: Production and accountability
- Advancedabout 24 hours6 stepsA secure enterprise copilotExtend your RAG assistant with sign-on, per-user document permissions, personal data masking and an audit trail, and write the security and compliance note a client's reviewers would ask for.Forward Deployed Engineer (FDE): Enterprise integration and governance
- Advancedabout 26 hours6 stepsA policy assistant people trustBuild a RAG assistant over a public document set of at least 300 pages, with citations, permission filters and an eval report comparing three retrieval set-ups, and show it refusing correctly.AI Engineer: Building with language models