Building: Privacy-first AI products (HIPPA, COPPA-compliant voice AI, hybrid RAG systems)

Expertise: Agent orchestration • Privacy-first AI • Voice AI • Data systems at massive scale (150TB/day)

Experience: 20+ years at Apple Maps, Salesforce, Microsoft building AI/ML platforms and data systems at scale

Helping Parents Nuture Curious Children. Living Trajectory System + Proactive Tracking

• AgentStellar Interactive AI learning platform with guided courses
• AgentMa Agent Observability and Management
• SessionStellar Sessions are the new Resume. Measure, Learn and Assess
• WowByDesign Collection of design skills for humans & AI agents

• Meduler Medication scheduling and reminder app

• NourishStellar Nutrition tracking and guidance
• TravelStellar AI-powered travel planning and booking

• ResumeStellar AI-powered resume tailoring and generation
• DecadeDirector Plan your life in decades
• MinusOneLabs Solo-founder product studio
• TreeTalk Record and store your family stories and health history privately

• AI Skills AI skills to help you with any AI agent
• claude-starter-kit Your AI Team Ready To Go
• ContextStellar AI prompt optimizer that cuts token waste instantly
• MCRS Medication Consensus Resolution Squad (MCRS) Data Specification
• ORFS Open Restaurant Feed Specification
• SPEAR Secure Prompt Enforcement At Runtime
• Voice101.ai www.voice101.ai Get started with Voice AI fundamentals

• Led 3 engineering teams (10 FTEs + 20 contractors) processing 150TB/day from 1000+ providers
• Launched world's first 3D Transit Maps, beating Google to market
• Shipped ML models to production: POI Hours prediction, Closure detection (75-95% precision)
• WWDC-featured work: Led Transit+Apple Pay, EV routing (iOS 13-16 tentpole Apple maps releases)
• Improved data pipeline throughput 50% through Spark optimization
• Contributed to Maps growth: 200M → 800M users

• Built production ML fraud detection processing 20TB at 1-2M events/sec
• Optimized Spark streaming performance 30% through custom partitioning
• Co-authored Security AI patent (USPTO) for deep learning anomaly detection
• Saved clients millions annually through fraud prevention

• Founded Bing Travel team (0→5 engineers), shipped cross-platform apps in 12 months
• Led data quality initiatives for MSN portfolio (400M users)
• Improved app performance 25% and user satisfaction 10%
• Authored ThinkWeek paper reviewed by President Qi Lu

Agent Orchestration:
  - Multi-provider abstraction (OpenAI, Claude, Gemini)
  - Tool registration with approval gates
  - Budget constraints and token tracking
  - Streaming with async generators

Memory Systems:
  - DDD architecture with bounded contexts
  - Scoped memory (session, user, global)
  - Conflict resolution strategies
  - Hybrid vector+keyword search
  - Memory outbox pattern (async ops)

Prompt Engineering:
  - Version-controlled in PostgreSQL
  - Evaluation pipeline with golden tests
  - Optimization: manual → DSPy → RLHF
  - A/B testing infrastructure

Safety & Compliance:
  - Red-team evaluation (promptfoo)
  - Multilingual attack corpora
  - COPPA compliance by design
  - PII tokenization and air-gap
  - Retention policies and audit trails

Voice AI:
  - ElevenLabs conversational agents
  - Twilio voice infrastructure
  - <1s latency optimization
  - Turn-taking, barge-in, VAD
  - Session memory across calls

Languages: TypeScript, Python, Java, Scala, Swift, C#, SQL
Data/ML: Spark, Kafka, Hadoop, Snowflake, TensorFlow, PyTorch, scikit-learn
Infrastructure: AWS, Kubernetes, Docker, Jenkins, Git, Streaming Pipelines
Databases: PostgreSQL, Cassandra, Redis, pgvector
Frontend: React, React Native, Next.js, Tailwind CSS

• Multi-agent systems with provider abstraction
• Tool execution with idempotency and approval gates
• Budget management and token tracking
• Streaming responses with backpressure handling
• Observable: distributed tracing, correlation IDs, structured logging

• DDD-based architecture with bounded contexts
• Scoped memory (session, user, global)
• Conflict resolution (LatestWins, HighestConfidence)
• Hybrid vector+keyword search (<50ms for 20M+ entries)
• Memory lifecycle: embedding, summarization, decay
• PII-aware with automatic redaction

• Promptfoo-based evaluation harness
• Multilingual attack corpora (4+ languages)
• Pre/post turn filters with COPPA mode
• Policy tiers (balanced, permissive, safe)
• CI gates: ≤0.1% leak rate, ≤2% false positives
• Prompt injection defense at scale

• LLM air-gap pattern (PII never reaches model)
• COPPA compliance: consent, retention, parental controls
• Data provenance: BASE/OVERLAY/DERIVED tiers
• Field-level licensing and expiration
• Automatic PII tagging and tokenization
• Audit trails for regulated domains

• ElevenLabs + Twilio integration
• <1s latency across full pipeline (STT → LLM → TTS)
• Conversational flows: greeting, turn-taking, barge-in
• Session memory across multiple calls
• Age-appropriate UX design
• Multi-channel: voice, SMS, web, devices

• 150TB/day ingestion from 1000+ providers (Apple)
• Entity resolution across 100M+ records
• Real-time streaming: 1-2M events/sec (Salesforce)
• ML productionization: evaluation, inference, monitoring
• Hybrid search: vector + structured filters
• Knowledge graphs with cross-domain traversal

Building world class products with an incredible team:

✅ AI/LLM products: Voice assistants, RAG systems, content moderation, regulated domains
✅ Privacy-first architecture: COPPA, GDPR, HIPAA compliance from scratch
✅ Data/ML systems: Pipelines, quality, entity resolution, forecasting, knowledge graphs
✅ Complex unblocking: Senior IC ownership with data + ML + infra fluency

Target domains:

• Children's products (COPPA expertise)
• Healthcare (privacy-sensitive data)
• Consumer AI products (voice, agents, RAG)

🏆 WWDC Featured: Transit+Apple Pay integration (2022)
📄 Patent: Deep Learning for Security AI (USPTO/WIPO, Salesforce)
📝 ThinkWeek Paper: "Bing Pins" reviewed by President Qi Lu (Microsoft)
🚀 Competitive Win: First 3D Transit Maps, beat Google to market (2021)
👥 People Leadership: Led multiple engineering teams to ship world's first 3D Transit maps 📈 Impact: Contributed to Maps growth 200M → 800M users

Best for:

• AI/LLM product development (0→1 or scaling)
• Production-grade agent orchestration systems
• Privacy-first architecture (COPPA, GDPR, HIPAA)
• Data/ML systems at scale
• Voice AI and conversational interfaces
• AI safety and evaluation
• Complex technical unblocking (senior IC)

"Building privacy-first AI products that respect users while delivering real value."