Projects • Prototypes • Experiments
I build things to answer questions research alone can’t.
A method, concept, and artifact used to test and demonstrate my work.
Tools used along the way
Instruments, not qualifications.
Evaluation Harness for a Research Repository Agent
A system for making retrieval quality measurable, reproducible, and debatable.
How it works
1. Model the repository
Research team models the scope, topics, and intent of the repository.
2. AI generates questions
AI generates diverse questions based on the modeling.
3. Assign & distribute
Each researcher is assigned questions from the evaluation set, creating a structured process for assessing outputs and operationalizing what good looks like.
4. Answer assigned set
Team members answer their assigned questions.
5. Cross-check & debate
Team reviews each other's answers, flags gaps, and debates where needed.
6. Refine & iterate
Consensus builds. We refine the model and improve over time.
What I built
- 10 inquiry types
- Evaluation test set
- Cross-checked answer key
- Debate workflow
What I test
- Retrieval quality
- Source grounding
- Guardrail behavior
- Human cross-validation
- Consistency & clarity
Impact
Makes AI quality measurable, reproducible, and open to structured debate — so the model can be improved iteratively rather than argued about.
Decision Intelligence Layer
Governed context assembly using MCP servers
A system concept for product teams, product and organizational leaders, researchers, and cross-functional groups who need the right context assembled before asking an AI system to support a decision. This is a concept in development, not a deployed production system.
How the concept works
1. Start with a problem
A user begins with a problem or a decision they need to make.
2. Identify sources
The system identifies potentially relevant enterprise sources.
3. MCP access
Model Context Protocol servers provide structured access to those sources.
4. Honor permissions
Existing permissions and access controls are honored.
5. Human confirms
A human confirms which sources are actually relevant.
6. Assemble context
Approved context is assembled for the model.
7. Decision support
The model provides analysis, options, tradeoffs, and next steps.
The goal
Improve decision quality by improving the context before the answer.
Who it's for
- Product teams shaping direction
- Product and organizational leaders making tradeoffs
- Researchers assembling evidence
- Cross-functional teams deciding together
Status
A concept in development. No implementation results, deployment, or performance claims are attached to it yet.
Designing and Shipping This Portfolio with AI
This website, treated as a project: one person using AI across an entire product workflow while keeping authorship and judgment.
Where AI helped
- Information architecture and site structure
- Interaction and layout exploration
- Visual prototyping
- Photo enhancement
- Writing development and refinement
- Quality review and iteration
What stayed with me
- Professional positioning
- Requirements and priorities
- Selection among generated options
- Factual accuracy
- Voice, tone, and ethical judgment
- Accessibility and final publishing decisions
Takeaway
AI increased the range and speed of the work, but maintaining a coherent point of view required deliberate human evaluation at every stage.