Skills demonstrated through building.

• AI-first UX design

  Designing interfaces where AI leads the experience: recommendations before controls, progressive trust, human-in-the-loop patterns.

  • ·3-level AI-first framework: L1 Recommend, L2 Adaptive UI, L3 Autonomous Execution
  • ·AI Triage as default landing, with narrative briefing before manual backlog
  • ·Team Creation Wizard: AI pre-fills everything, humans override
  • ·Deploy Risk Assessment: visualized risk scoring before destructive actions

• Specification precision

  Writing instructions so explicit that machines reproduce your exact intent, leaving no gaps for hallucination to fill.

  • ·21 agent specs with explicit capabilities, constraints, and output schemas
  • ·Per-agent tool scoping via toolsEnabled arrays (principle of least privilege)
  • ·Spec Pipeline: 4-level maturity system with LLM enrichment for agent-ready stories

• Task decomposition

  Breaking complex objectives into agent-sized tasks. Knowing which subtasks an agent handles solo vs. which need human oversight.

  • ·5-stage intelligence pipeline: Discovery, Analyst, Prototype, Persona, Product Owner
  • ·Pointer-based handoff protocol with 40x token reduction vs. full-document passing
  • ·4 team topologies: Solo, Supervised, Collaborative, Hybrid
  • ·Agent execution pipeline: story, worktree, implement, test, PR, with clear handoff points

• Adoption-aware design

  Designing AI products for the user who actually shows up, not the one the builder wishes existed. Serving teams across the adoption curve in one product.

  • ·Dual-mode workflow: Today (daily cycle, agent-native) and Work (sprint-based) on shared data and AI infrastructure
  • ·Company-level work-mode preference (agent / dual / human) swaps surfaces at runtime. No migration required.
  • ·Same artifact schema underneath, so spec-enrichment, smart-assign, and agent execution work in either mode

• Context architecture

  Organizing knowledge so AI agents find exactly what they need. The "Dewey Decimal System" for your AI, covering persistent vs. ephemeral context.

  • ·Context Engine: getRelevantContext(), getRecommendationContext(), getStoryContext()
  • ·Tiered onboarding: 5 min (~70% AI effectiveness) to 30 min (~95%)
  • ·KB Quality Score: coverage assessment across 7 categories with gap identification
  • ·Scoped retrieval per feature, so consumer chat vs. internal agents see different context

• RAG pipeline design

  End-to-end retrieval-augmented generation: embedding strategy, hybrid search, prompt assembly, and output grounding.

  • ·Full embedding pipeline with pgvector, stale marking, and re-embedding triggers
  • ·Hybrid search: full-text + vector similarity with configurable weighting
  • ·Production-deployed Consumer Chat: external-facing RAG chatbot with scoped retrieval
  • ·Knowledge Chat: multi-turn RAG conversation with source attribution

• Multi-agent orchestration

  Designing how multiple AI agents collaborate: handoff protocols, shared state, error propagation, and recovery strategies.

  • ·AgentWeave: purpose-built orchestration engine with pipeline, solo, and collaborative modes
  • ·21 agents across 7 teams with defined topologies and supervisor styles
  • ·Training Arena: A/B testing different orchestration configurations against benchmarks
  • ·Team config export to Claude Code, Cursor, GitHub Copilot, Aider, and Continue.dev

• Prompt engineering at scale

  Moving beyond one-off prompts to systematic prompt management: database-stored, admin-editable, multi-provider compatible.

  • ·21 agent prompts stored in database, editable from admin UI, not hardcoded
  • ·Model config table: prompts paired with per-feature model selection
  • ·Multi-provider compatibility: same patterns work across Anthropic Claude and Google Gemini
  • ·Structured output schemas: agents return typed JSON or Markdown with specific formats

• Evaluation & quality

  Building systematic ways to measure whether AI output is actually correct, not just fluent. Catching the subtle failures.

  • ·Model Calibration: golden test cases with side-by-side comparison across providers
  • ·Training Arena: LLM-as-judge evaluation scoring completeness, precision, feasibility
  • ·Inference Inspector: full retrieval attribution to verify source relevance
  • ·Agent defect tracking with auto-creation on build/test failure

• Trust & security

  Designing boundaries between human and machine. Building blast radius assessments. Ensuring AI systems fail safely.

  • ·8-scanner security pipeline: npm audit, auth guards, SQL injection, XSS, RLS, and more
  • ·Git worktree isolation for agent execution: disposable branches, never main
  • ·Per-agent tool scoping, team guardrails, and configurable review policies
  • ·RLS write-blocking on all 85+ tables for demo user, belt-and-suspenders protection

• Cost & token economics

  Calculating ROI of every AI operation. Right model for the right task. The new infrastructure cost optimization.

  • ·3-tier model config: Lite (Haiku/Flash), Standard (Sonnet), Advanced (Opus) per feature
  • ·Budget service: monthly spend tracking, burn rate, projection, stories-remaining
  • ·Per-agent token tracking logged to every run with cost estimation
  • ·AI Triage surfaces estimated token cost per recommended action

• AI observability

  Monitoring AI systems in production: not just uptime, but output quality, cost trends, and retrieval attribution.

  • ·Inference Inspector: per-trace observability with plain-English explanations
  • ·Workforce Monitor: real-time activity, queue depth, token burn rate
  • ·Agent run logging: full prompt + response + metadata captured per execution
  • ·Security scanner with daily automated runs, auto-assignment, and auto-resolution

Product decisions are more valuable than features. Each of these was a deliberate choice with a specific rationale and measurable outcome.

• Pointer-based context passing

  WhyFull-document passing bloats context windows and wastes tokens. Agents pass only UUIDs and summaries, forcing clean boundaries.

  Outcome40x token reduction in multi-agent pipelines while maintaining output quality.

• Context investment as the AI quality lever

  WhyAI effectiveness isn't a model problem. It's a context problem. The more product context the system has, the better its outputs. But most teams don't measure or budget for it.

  OutcomeBuilt as Product Context Setup, a tiered onboarding flow where each tier maps to a measurable AI quality target: a 5-minute Quick Start reaches ~70% effectiveness, a 30-minute full walkthrough reaches ~95%.

• Recommendations before controls

  WhyTraditional tools show a wall of data and expect humans to find the signal. AI-first means the system does the analysis and presents actions.

  OutcomeAI Triage as default Work page, delivering a narrative briefing with cost estimates, risk, and recommended actions.

• Anti-platform architecture (fork, don't rent)

  WhyMulti-tenant SaaS locks customers into shared infrastructure with no customization. Each company should own their instance.

  OutcomeModule manifest + schema split + config-driven agents, customizable without code changes.

• Dual-mode workflow on shared infrastructure

  WhyCompanies sit at different points on the AI adoption curve. A product that only serves the destination fails the team using it today.

  OutcomeToday + Work modules coexist on the same data model, selectable per company. Traditional teams adopt AI-as-assist; agent-native teams run daily cycles; hybrid teams use both.