JJoeven

Projects

Each project is a sequence: architecture, environment, loop, tests, hardening. Ship them as a public portfolio.

01 · beginner · 5–7 hours

Weather Tool Agent

Build a user-goal agent with geocode, weather, and finish tools, a JSON action protocol, retries, and tests — all simulated in stdlib Python.

You will ship: A runnable weather agent that geocodes a city, fetches a forecast, finishes with a sentence, retries bad JSON, and fails closed on unknown places.

tool registriesJSON actionsretriesgoal predicatesunit tests

02 · intermediate · 6–8 hours

ReAct Research Agent

Implement a ReAct loop over a fake web corpus: thought, search, citations, max steps, and a cannot-answer path when evidence is missing.

You will ship: A research agent that searches a toy corpus, quotes sources, stops at a step cap, and says it cannot answer when the corpus is silent.

ReActsearch toolscitationsmax-step budgetsabstention

03 · intermediate · 6–8 hours

RAG Customer Support Agent

Chunk a tiny product handbook, retrieve with cosine similarity over bag-of-words vectors, answer with citations, and refuse when retrieval is weak.

You will ship: A support agent that answers only from retrieved handbook chunks, cites chunk ids, and refuses when similarity is below a threshold.

chunkingbag-of-words vectorscosine similaritycited answersrefusal thresholds

04 · advanced · 8–12 hours

Multi-Agent Software Team

Run planner, coder, and reviewer agents over a shared repo dict until fake tests pass — with budgets, patches, and no infinite rewrite loops.

You will ship: A three-agent coding team that plans, edits a dict-as-repo, reviews, and stops when a fake test runner is green — or when the budget dies.

multi-agent rolesshared statepatchestest oraclesstop conditions

05 · advanced · 8–12 hours

Autonomous Ops Agent

Monitor fake metrics, diagnose incidents, propose actions, require HUMAN_APPROVAL before anything destructive, and write an incident report.

You will ship: An ops agent that detects SLO burn, proposes a bounded action plan, blocks destructive tools without HUMAN_APPROVAL, and emits a post-incident report.

metrics monitoringdiagnosisaction proposalshuman approval gatesincident reports