Workspace/Capstone studio
Loading progress
BUILD SOMETHING YOU CAN DEFEND

Capstone studio

Eight research-oriented directions. Choose one, narrow the question, and build convincing evidence.

A strong portfolio starts
with a testable question.

Each brief includes a working seed, system design, baselines, evaluation metrics, milestones, a demo script, and an honest resume template.

Original proposals, open questions.These are opportunities to investigate. Novelty requires a related-work review; impact requires your measured results.
Industrial telemetry, typed ontologies, and engineering digital twins

Evidence Twin

An industrial cooling investigator that knows when its evidence is insufficient

Research6-week suggested build
3 baselines · 3 experimentsExplore blueprint
Agent security, reliability evaluation, and tool boundaries

ToolTrust

A wind tunnel for agents exposed to stale, contradictory, and malicious tool results

Research6-week suggested build
3 baselines · 3 experimentsExplore blueprint
Site reliability engineering, counterfactual planning, and recovery

Incident Rehearsal

An incident agent that rehearses candidate fixes and proves its rollback path inside a simulator

Research6-week suggested build
3 baselines · 3 experimentsExplore blueprint
Temporal retrieval, consent-scoped memory, and deletion evaluation

Forgetful Memory Lab

A benchmark for remembering the right version and deleting its downstream traces

Research6-week suggested build
3 baselines · 3 experimentsExplore blueprint
Accessibility, browser agents, and human-agent interaction

AccessBridge

A verifiable assistant for navigating synthetic forms with accessible state and explicit confirmation

Advanced6-week suggested build
3 baselines · 3 experimentsExplore blueprint
Research agents, claim graphs, and temporal evidence retrieval

Living Evidence Observatory

A research monitor that preserves disagreement, provenance, and what was known at the time

Research6-week suggested build
3 baselines · 3 experimentsExplore blueprint
Coding agents, reusable skills, evaluation, and capability control

Skill Regression Foundry

A release laboratory for versioned Markdown skills with behavioral and privilege regression gates

Advanced6-week suggested build
3 baselines · 3 experimentsExplore blueprint
Agent protocols, distributed state, and compatibility testing

Protocol Clinic

An interoperability simulator that localizes schema and lifecycle drift across MCP, A2A, and AG-UI

Research6-week suggested build
3 baselines · 3 experimentsExplore blueprint

Suggested durations assume focused project time after the relevant modules. No project title alone guarantees differentiation or a job. Your implementation quality, experiments, and explanation make it valuable.