I design and lead reliable AI systems for enterprise use—where model behavior, retrieval, orchestration, integrations, guardrails, and production operations have to work as one system.
Portfolio · LinkedIn · Résumé · Technical writing
| 200K+ | 100K+ | 1 of 4 | 100+ |
|---|---|---|---|
| Employees supported by the enterprise platform | Monthly conversational AI interactions | Development leads across four teams | Merged open-source pull requests |
I serve as a development lead for an enterprise conversational AI platform at Bank of America. My work spans GenAI and RAG architecture, LLM integration and evaluation, agentic orchestration, API and integration design, production reliability, and engineer mentorship. Public résumé results include 30%+ fewer live escalations per topic and 20%+ fewer incident tickets.
Enterprise work is intentionally described at a public-safe level. Internal architecture, code, systems, and data are not shared here.
- Agentic orchestration: routing policies that choose deterministic workflows, retrieval, clarification, tools, or human escalation with explicit controls.
- Enterprise RAG: governed evidence, retrieval quality, evaluation sets, confidence, latency, and measurable failure modes.
- Conversational systems: intent, entity, context, and dialogue patterns that turn ambiguous language into safe, maintainable journeys.
- Integrations and platforms: typed API contracts, authentication, workflow automation, failure handling, and reusable enterprise patterns.
- Guardrails and reliability: validation, fallbacks, access boundaries, observability, SLOs, staged delivery, and production-readiness reviews.
| Project | What it demonstrates | Stack |
|---|---|---|
| Enterprise Conversational AI | Technical leadership for retrieval, routing, integrations, controls, escalation, and production readiness at enterprise scale | GenAI · RAG · Agentic orchestration · APIs · Observability |
| CleanShare Pro | Review-first detection and redaction of sensitive information in images and documents | Next.js · TypeScript · Capacitor · WebAssembly · OCR/PII |
| AI Calendar | Natural-language scheduling translated into explicit, validated, reviewable actions with manual fallback | Groovy · JavaFX · Structured actions · REST |
| CodeCap | Desktop capture, OCR, review, categorization, and local search for code and technical notes | Electron · TypeScript · OCR · SQLite |
| Computer Vision Safety Pipeline | Object and activity detection with threshold tuning and post-processing to reduce false positives | Python · PyTorch · YOLOv5 · SlowFast |
More architecture detail and public-safe case studies are available in the portfolio.
I have 100+ merged pull requests overall, including changes accepted by projects across Meta, NVIDIA, Microsoft, Google, and Firebase. The work spans features, bug fixes, core refactors, developer tooling, CI and release safety, tests, and documentation.
| Project | Contribution | Type |
|---|---|---|
| Meta · Lexical | Consolidated duplicated ancestor-lookup logic into a shared typed utility | Core refactor |
| NVIDIA · cuCollections | Added a CUDA heterogeneous-lookup example for static_map |
Developer education |
| Microsoft · MsQuic | Added a CI size gate to protect Rust crate publishing | CI / release safety |
| Microsoft · VS Code Mypy | Restored diagnostics by parsing both stderr and stdout | Bug fix |
| Google · ML Compiler Opt | Clarified TensorFlow policy-saver usage and type intent | Maintainability |
| Firebase · FlutterFire | Corrected Android OAuth client documentation | Documentation |
| Google Cloud · Gemini Cloud Assist MCP | Corrected CLI guidance in the project documentation | Documentation |
- Designing a Reliable Enterprise AI Router — a control-plane pattern for deterministic workflows, retrieval, clarification, and human escalation.
- Improving RAG Accuracy Without Ignoring Latency — treating retrieval, evidence, model choice, latency, and evaluation as one production system.
- AI and architecture: GenAI, RAG, LLM integration/evaluation, agentic orchestration, MCP, conversational AI, AI guardrails, NLP/NLU, computer vision
- Software and integration: Python, PyTorch, Groovy/Java, TypeScript/Node.js, C#, REST APIs, Apache Camel, authentication, ServiceNow integrations
- Reliability and delivery: CI/CD, GitHub Actions, Azure DevOps, Jenkins, Docker, Terraform, logs/metrics/traces, SLO/SLI, incident response
- Cloud and data: Azure, AWS, GCP, PostgreSQL, MS SQL, Supabase, Firebase, vector retrieval, caching, and queues
I write about dependable AI architecture and build privacy and developer tools end to end. The best starting points are my portfolio, LinkedIn, or email.