WORK & PORTFOLIOWork & portfolio: a path from ideas to working systems.
These projects do not claim that AI can do everything. They record how Frank learns, orchestrates AI, verifies results and keeps refining across different problems—leaving methods that can be taught.
PRODUCTKNOWLEDGEWORKFLOWDATASYNTHESISCASE 01 · PRODUCT · ACTIVEPremiumPets
The original idea
Organize pet services, knowledge, commerce and community into a genuinely useful product—not merely an app that opens.
What the journey revealed
After roughly four months of outsourcing, the prototype remained far from the intended product. With AI, Frank entered product, data, design, code, security review and iteration himself.
Method carried forward
Protect the real problem, decompose roles, orchestrate AI, verify output and revise until the work is responsible to both maker and user.
CASE 02 · AI WORKFLOW · EVOLVINGAI Control Studio
Why it exists
When one project needs research, product, design, code and testing, a single chat box cannot preserve roles, context, handoffs and ownership.
How it works
Different AIs, task briefs, inputs, outputs, human review and next steps are organized in one flow. Frank keeps the goal, delegation, trade-offs and final acceptance.
How it enters teaching
Learning Studio grows from this practice: students see not only answers, but how a whole project moves.
MORE THAN ONE PROJECTThe same core capability, practised across very different problems.
The interfaces and scale differ, but the path is consistent: define the problem, organize knowledge, orchestrate AI, return to evidence and keep human judgment.
03QUERY / RETRIEVE / CITE
LOCAL AI INFRASTRUCTURELocal RAG System
Practised · evolving- PROBLEM
- Large collections lived on a computer but remained hard to find, verify and reuse.
- PRACTICE
- Local parsing, chunking, indexing, retrieval and source tracing let AI answers return to evidence instead of sounding merely plausible.
- LEARNING
- Useful AI must show what it relied on and let people judge the source.
0457 → 44,455 → ARTICLE
DOMAIN KNOWLEDGEPet Encyclopedia Knowledge Base
MVP complete · content evolving- PROBLEM
- Pet knowledge is fragmented and uneven; ordinary search rarely balances depth, clarity and sources.
- PRACTICE
- 57 sources became 44,455 RAG records, a cat-breed sub-library, a first Ragdoll entry and an H5 MVP.
- LEARNING
- Page structure is only a container; evidence, depth, expression and updating make the knowledge product.
05SOURCE / INDEX / TRUTH
PERSONAL KNOWLEDGE OSLong-term Memory & Truth Vault
In operation- PROBLEM
- Across AIs, threads and long projects, decisions are easily lost and old summaries can override current truth.
- PRACTICE
- Original conversations, decisions, indexes and machine retrieval are layered so claims remain traceable and AI can distinguish clues from final truth.
- LEARNING
- More memory is not automatically better; provenance, priority, freshness and conflict rules matter.
06CRAWL / CLEAN / COMPARE
DATA PIPELINECatalog & Price Benchmark Pipeline
Repeatable workflow established- PROBLEM
- Supplier catalogs, market prices and specifications come from different sources; manual matching is slow and inconsistent.
- PRACTICE
- AI helps collect, clean, deduplicate, normalize, check anomalies and compare by rules; people confirm sources, matches and commercial judgment.
- LEARNING
- Automation removes repetition, but data definitions and decisions remain human responsibilities.
07SOURCE / STORY / ACTION
AI SYNTHESISAI Brief & Proposal Generation
Experimented and used- PROBLEM
- Having many sources and ideas does not automatically create a clear, credible and actionable brief.
- PRACTICE
- AI learns sources, extracts facts, shapes outlines, narrative and visual hierarchy; human verification and revision turn them into project briefs, design notes and reports.
- LEARNING
- A good brief does more than shorten text: it knows what matters, what is evidenced and what action should follow.
A portfolio is not a cabinet of outcomes. It is a map of learning.
01Idea & problem
Why it began and whom it should help
02Decompose & orchestrate
What AI carries and what judgment stays human
03Evidence & revision
Where AI failed and how people changed the result
04Delivery & reflection
Whether it solved the problem and what comes next
STUDENT PORTFOLIO · BEGINS WITH COHORT 01Student work will be documented along the same evidence chain.
Original problem and interviewsVersions and key decisionsAI contribution and human revisionTesting and acceptanceStudent reflection (with consent)Family and partner feedback (with consent)
The first cohort has not begun, so no invented student work or reviews appear here. Real work will join after completion, privacy review and consent.
See the first student project 
