Portfolio project briefs for practical AI work
These are original public briefs derived from the capabilities taught in Ropes. They are not student submissions, client case studies, or promises of commercial results.
A small sales team reviews every inbound lead manually.
Map qualification rules, add source-grounded enrichment, use AI only for ambiguous classification, and require approval before routing.
Workflow map, configured automation, evaluation cases, decision log, and handover guide.
Employees cannot find consistent answers across policies and SOPs.
Prepare a safe document set, design retrieval, require citations, test unanswered questions, and state access assumptions.
Working RAG assistant, retrieval test set, limitations report, and operating documentation.
A tool-using assistant has been built without a structured threat review.
Map assets and trust boundaries, test injection and permission abuse, apply layered controls, and document residual risk.
Threat model, red-team cases, mitigation evidence, and review report.
Prompt and model changes ship without evidence that quality remains acceptable.
Create a golden dataset, define task-specific criteria, benchmark a baseline, and add release thresholds.
Evaluation harness, results dashboard, regression policy, and findings summary.
A useful model exists only as an untracked notebook.
Track experiments, package the model, expose an endpoint, monitor performance, and define retraining triggers.
Reproducible pipeline, container, endpoint, monitoring view, and runbook.
A team wants an AI feature but has not defined value, failure behaviour, or success criteria.
Research the user need, frame the AI role, define non-AI fallbacks, set evaluation metrics, and document trade-offs.
Feature brief, PRD, evaluation plan, risk register, and two-quarter roadmap.