I design and build software around difficult human workflows - then add the state, testing, governance, provenance, accessibility, and human-control layers required to make those systems trustworthy.
Portfolio boundary: public-safe, synthetic, de-identified examples only. No participant, patient, family, child, or private legal records are displayed.
PRODUCTSFrontline workflow software
AI & SYSTEMSGoverned model and agent control
INFRASTRUCTUREContinuity, provenance, validation
FEATURED WORK / 01-06
Six systems. One thesis.
The front page is intentionally curated. The Atlas underneath it preserves the wider body of software, specifications, prototypes, infrastructure, and historical work without pretending every item has the same maturity.
01NON-PRODUCTION PROTOTYPE
Treatment Signal
Behavioral-health workflow and treatment coordination
Operational continuity for treatment-related work: ownership, handoffs, verification, human decisions, auditability, and minimum-necessary information - without pretending to replace the clinical source of truth.
Product
Healthcare
Workflow states
Supabase/Postgres
RLS
Human review
Product decision
Separate operational coordination from clinical authority. The system can coordinate status and next actions while preserving a hard boundary around clinical decision-making and source-of-truth records.
Evidence boundary
Independent synthetic/non-production prototype. No production PHI, production staff accounts, or live clinical integrations are represented here.
02FUNCTIONING SYNTHETIC PROTOTYPE
Smart Family Flow
Family needs, referral ownership, follow-up, and QR intake
A two-sided workflow: families ask for help in plain language; staff receive a structured Need that can be claimed, routed, worked, followed up, resolved, and audited.
Product
Human services
Next.js
TypeScript
Command model
QR intake
Design principle
Figure out what the person needs before showing them who handles it. The client should not need to understand an organization's internal chart to ask for help.
Boundary
Independent synthetic prototype informed by frontline service workflows. It is not represented as commissioned, adopted, or endorsed by an institution.
03TESTED PROTOTYPE
Workforce Continuity
Frontline handoffs, consent, next actions, and reviewer feedback
A mobile-first continuity layer for high-friction frontline workflows, with explicit consent behavior, persistence boundaries, accessible interaction patterns, and structured human feedback receipts.
Product
Workforce
HTML/CSS/JS
Accessibility
Consent
Static tests
What changed through iteration
The prototype evolved into a reviewable workforce concept with mobile-safe feedback capture, share/copy fallbacks, privacy reminders, and explicit non-persistence of unsent reviewer notes.
Boundary
This portfolio presents the generalized workforce system only. No employer commissioning, adoption, or endorsement is claimed.
04PARTIAL RUNTIME / ACTIVE BUILD
CourtChain
Governed legal matter continuity without autonomous legal authority
A matter-oriented system for timelines, evidence, deadlines, filing states, accessibility routing, provenance, authentication, and fail-closed access boundaries.
Product
Legal
Accessibility
Auth/session
CI gates
Provenance
Safety boundary
The software is designed to preserve legal information, deadlines, provenance, and workflow state while keeping legal authority and actual filing decisions with authorized humans.
Portfolio privacy
Only synthetic matter structures are appropriate for public demonstration. Private case evidence and strategy are excluded.
05REFERENCE IMPLEMENTATION
HCOS Governed AI Execution
Provider-independent AI execution and authority controls
A model-independent control pattern for scoped context, adapter routing, authorization, fallback behavior, provenance, schema validation, and human authority.
AI infrastructure
Governance
Python
JSON Schema
Fail-closed policy
Conformance tests
Core rule
A model can generate, propose, summarize, or evaluate. It does not silently inherit authority merely because it can communicate with another model, agent, or tool.
Architecture goal
Keep identity, durable authority, and governed continuity outside any single model provider so models remain replaceable execution components.
06ACTIVE ENGINEERING SYSTEM
Human Code Continuity Spine
Cross-repository provenance, convergence, and project continuity
A repository-backed system for harvesting source context, preserving provenance, classifying conflicts, routing records, validating machine-readable contracts, and preventing chat memory from silently becoming authoritative project state.
Long-running AI-assisted programs accumulate conversations, branches, decisions, superseded designs, and partially merged work. The spine preserves source lineage and forces an explicit distinction between evidence, disposition, and accepted governing state.
Current boundary
The convergence branch remains an active draft/harvest system. It is not represented as a completed merge or production transition.
See software, AI systems, infrastructure, protected work, and concept history with explicit maturity labels.
PLATFORM CORE / HUMAN-CONTROLLED CONTINUITY
Continuity Wallet is where the architecture becomes a product.
Identity, consent, contextual roles, app routing, continuity history, guided packet preparation, and audit behavior come together in one human-facing command center. The Wallet is not “just another project”; it is a connective product layer underneath multiple Human Code experiences.
Maturity boundary: the public Wallet surface is inspectable, but that does not mean every connector, database path, authentication journey, or enterprise admission gate is production-complete.
IDENTITYwho is acting
CONSENTwhat may move
ROLEcontextual authority
ROUTEapproved workspace
RECEIPTwhat happened
EXPORTportable continuity
INSPECTABLE / ACCESS-GATED SURFACES
Don’t just read about the software.
Where a public-safe route exists, open it. These links carry different maturity and access states; reachability is not represented as universal production readiness.
Recruiters should not have to infer the technical body of work from project names. This index makes the build surface explicit while the Atlas carries the evidence and maturity detail.
The strongest systems work starts before the code: who is confused, what gets dropped, where authority is ambiguous, what data should not move, and what must remain true after a handoff or model call?
01
Observe the human workflow
Find friction, ambiguity, handoff loss, access barriers, and repeated work.
02
Model the state
Define objects, roles, transitions, ownership, evidence, and failure paths.
03
Build the smallest proof
Prototype the critical vertical slice with synthetic data and explicit boundaries.
04
Test behavior, not screenshots
Exercise invalid transitions, refresh, fallback, accessibility, isolation, and regression.
05
Keep humans authoritative
Approval, revocation, provenance, privacy, and action receipts are product behavior.
06
Ship evidence with the change
Code + tests + documentation + specification, with preview-first promotion.
THE THESIS
Technology should preserve the continuity of the human using it.
My path into product and AI started from the human side of fragmented systems: family services, behavioral health, homelessness services, court navigation, accessibility, and high-context coordination.
That background now informs how I design software: fewer hidden handoffs, clearer ownership, stronger evidence, explicit boundaries, and AI that supports - rather than quietly replaces - human judgment.