Work & Evidence
The work, and what’s actually verifiable.
Doctrine is easy to claim; work is harder to show. This page is the evidence chain — a proof table, the projects, a few case studies, and an honest label on every one: what’s public, what’s private, and what’s still being assembled. Same fact discipline I use everywhere else.
Verified public artifact you can open
Calculated deterministic result/metric
Approved reviewed by a named stakeholder
Private under NDA / on request
In progress being assembled
12+ miniapps, made concrete
A representative set. Items marked built at UST are professional work shown as experience — not personal products and not for sale.
| Miniapp | Sector | AI role | Status | Evidence |
|---|---|---|---|---|
| RevTXM | Revenue / SaaS | AI-assisted, deterministic core | Production (early) | Private walkthrough on request |
| ClarisTXM for Services | Home services / SMB | AI-forward workflows | In development | In progress |
| TXM Studio / DoReMini | App builder | Intent → governed microapp | Design / concept | In progress |
| Rapid Impact Analyzer | SAP (built at UST) | AI explains; engine scores | Live on SAP BTP | Private employer work |
| UST AI Suite+ for SAP | SAP (built at UST) | Governed AI portfolio | Live | Private employer work |
| IT Debt Analyzer | AMS / ITSM (built at UST) | AI on clusters only | Shipped prototype | Private employer work |
| + additional builds | CPG, SMB, consultants | Mixed | Varies | Private |
Fact disciplineCounts and outcomes are labeled, not inflated. Where a result isn’t independently verified yet, it says so.
Selected projects
RevTXM
Personal- Problem
- Revenue leaders drown in tool pitches and stalled AI pilots; they need a governed revenue system, not another dashboard.
- System
- Multi-tenant SaaS, subscription billing, deterministic core with provenance labels on every figure and approval gates on consequential actions.
- AI role
- AI generates the revenue architecture and ranks opportunities; it never computes the numbers.
- Outcome
- In production, early. In progress
PrivateWalkthrough on request
ClarisTXM for Services
Personal- Problem
- Home-service and SMB operators run their business across a dozen disconnected tools, from intake to billing.
- System
- An AI-forward operating system for the lead-to-repeat lifecycle — intake, quoting, scheduling, delivery, billing, and customer experience as governed miniapps.
- AI role
- AI drafts and assists inside each workflow; humans approve anything that touches money or the customer.
- Outcome
- In development. In progress
In progress
TXM Studio / DoReMini
Personal- Problem
- Teams want governed AI microapps but can’t turn business intent into safe, audit-ready execution systems.
- System
- A miniapp builder and execution workbench — a plugin contract with deterministic formulas, bounded AI actions, source-grounding, and an audit slot.
- AI role
- Bounded, named AI actions inside a fixed app boundary — the motor inside the machine.
- Outcome
- Design / concept, contract stress-tested. In progress
In progress
Rapid Impact Analyzer
Built at UST- Problem
- SAP teams can’t see change impact across custom code, transports, integrations, requirements, tests, and operational evidence.
- System
- A deterministic impact model with AI-assisted explanation, evidence ingestion, provenance labels, and an audit trail — live on SAP BTP Cloud Foundry.
- AI role
- AI explains, clusters, and drafts; the engine computes the score.
- Outcome
- Designed to compress a 6-week, 12-consultant assessment to a 2–4 hour governed session design target
PrivateEmployer work — shown as experience, not for sale
IT Debt Analyzer
Built at UST- Problem
- AMS ticket backlogs are opaque; manual classification of technical debt takes weeks.
- System
- Local, deterministic clustering across 17 debt taxonomies; the AI sees cluster summaries only, so raw client tickets never leave the machine.
- AI role
- AI names and reasons over clusters; the clustering and scoring stay deterministic.
- Outcome
- Shipped prototype, zero-install. Private
PrivateEmployer work
The Business AI Architect Method
Published IP- Problem
- There’s no shared discipline for deciding what to build with AI, where it belongs, and how it survives compliance.
- System
- An operating model — the cost ladder, provenance labels, approval gates, honest UI, fact discipline — published as the TXM Body of Knowledge.
- AI role
- The method governs where AI is used at all.
- Outcome
- Published and applied across the portfolio. Verified
VerifiedThe principles →
Case studies
Structured the same way every time: problem, approach, what the AI does, what stays deterministic, outcome, and evidence status.
SAP change-impact analysis Built at UST
- Problem
- SAP teams struggle to understand change impact across custom code, transports, integrations, requirements, test cases, and operational evidence.
- Approach
- A deterministic impact model with AI-assisted explanation, evidence ingestion, provenance labels, and an audit trail, deployed on SAP BTP.
- AI role
- Explains, clusters, summarizes, and drafts recommendations. It does not compute the score.
- Deterministic role
- Rules, relationship mapping, scoring logic, and approval status stay deterministic and reproducible.
- Outcome
- Design target: a 6-week, 12-consultant assessment compressed to a 2–4 hour governed session. In progress — independent client metrics not yet published.
- Evidence
- Private employer work; walk-through available under appropriate terms.
Revenue management as a governed system Personal
- Problem
- CROs are told to deploy AI across the revenue engine but have no trustworthy way to decide which tools and claims to believe.
- Approach
- RevTXM — a multi-tenant SaaS where AI proposes the revenue architecture and a deterministic core, provenance labels, and approval gates keep every figure defensible.
- AI role
- Generates the 5-view architecture and ranks opportunities; never computes the underlying numbers.
- Deterministic role
- The data layer, the math, and the billing logic are deterministic and auditable.
- Outcome
- In production, early. In progress — tenant/revenue figures shared on a walkthrough.
- Evidence
- Private — live walkthrough on request.
Deterministic-first AMS debt analysis Built at UST
- Problem
- A services org needed to see the technical debt buried in years of AMS tickets without sending client data to the cloud.
- Approach
- Local TF-IDF clustering across 17 debt taxonomies; the AI consults on cluster indices only, never raw rows; a regression checker validates against locked baselines.
- AI role
- Names clusters, estimates automation potential, drafts the reasoning pack.
- Deterministic role
- Parsing, clustering, scoring, and the hours-saved math stay deterministic and reproducible.
- Outcome
- Shipped as a zero-install prototype. In progress — client outcome metrics private.
- Evidence
- Private employer work.
GitHub / public proof-of-work
My public GitHub is being organized as a proof-of-work portfolio for applied AI architecture — sanitized prototypes, architecture references, product specs, runbooks, and demo miniapps. It’s honestly In progress; today most repos are private. Planned public repos:
- business-ai-architect-playbook — the method, the cost ladder, provenance labels, as an open spec
- txm-studio-concept — the miniapp plugin contract and a reference microapp
- btp-deploy-runbook — the Node-to-Cloud-Foundry deploy runbook (generic, vendor-neutral)
- ai-decision-tool — the deterministic AI-placement tool, open-sourced
Once each is live, it moves from planned to a Verified card here.
Deciding whether to hire, fund, or buy?
The role-fit and how-to-evaluate-me page is next door; the résumé is one click; or just email me and ask for whatever evidence you need.
How to hire me Résumé