AI Solution Architect & Full-Stack Engineer based in NYC.
I specialize in building autonomous AI agents and scalable web architectures. With a background as a 2x Founder, I bridge the gap between complex engineering and business impact.

🎙️ ARS — Adaptive Restaurant Speech An ASR engineering deep-dive (exploration, not a product)

• A hands-on study of a problem I find fascinating: bilingual (es + en) speech recognition in drive-thru noise — built as a self-improving pipeline (noise lab → LoRA fine-tuning → rule-based correction) to measure what actually helps, not assume it.
• Everything is gated: a Noise Damage Index ranks noise types by measured WER/KER damage, denoisers that sound cleaner but transcribe worse get auto-rejected, and every shortcut is documented in an honest build report — bugs included.
• Comes with design reflections on where systems like this should go: the acoustic front-end, the ASR→NLU handoff, and measuring user friction beyond WER.

🤖 Odoo AI Agent Production-Grade Voice-Enabled ERP Assistant → live demo

• Developed a 21-node LangGraph pipeline enabling complex natural language interaction with Odoo ERP modules (Sales, Inventory, Finance, CRM).
• Voice interface: Speech-to-Text (Groq Whisper) + Text-to-Speech (Kokoro ONNX) with synchronized streaming — talk to your ERP.
• Safe CRUD Operations: strict confirmation gates and interactive ambiguity resolution cards to ensure zero accidental data modifications.
• Data-Driven UX: SSE streaming for real-time responses, integrated live analytics, and automated Excel/PDF reporting engine.
• Architecture: Next.js 16 (React 19) + FastAPI + LangGraph + Odoo XML-RPC.

🎯 TalentGraph AI-Powered Voice Recruitment Platform → live demo

• Developed a 34-node LangGraph agent with a multi-level detection system for ultra-low latency (<10ms for Fast Paths).
• Multi-modal I/O: Integrated Speech-to-Text (Groq Whisper) and Text-to-Speech (Kokoro ONNX) with synchronized streaming.
• Autonomous Workflows: Implemented MCP integration for automated Google Calendar scheduling and Gmail management.
• Performance: Features Silent Loading and heuristic routing, achieving a 70% reduction in context-related latency.
• Architecture: Angular 19 (Signals) + FastAPI + LangGraph + PostgreSQL (Checkpoints).

📚 AI-Trainer-Teacher Intelligent RAG-powered Educational Platform → live demo

• Engineered a hierarchical RAG pipeline with dynamic page filtering, reducing irrelevant context by 70%.
• Implemented history-aware question reformulation and Redis-backed session persistence for multi-turn learning.
• Impact: Faster retrieval and higher precision by narrowing vector search scope to specific document sections.

Also on the shelf: LangGraph agent templates (6 production-ready agentic patterns) · OAuth 2.0 & JWT deep-dives (PKCE, Auth Code, real-world API integrations).

Speech & ML: faster-whisper (CTranslate2) • PEFT/LoRA fine-tuning • Silero VAD • STT/TTS streaming (Groq Whisper, Kokoro) • WER/KER evaluation pipelines
AI Engineering: LangGraph • LangChain • RAG Orchestration • Model Routing • Vector DBs (Pinecone/Chroma)
Frontend: Angular 19 (Signals/Standalone) • React (Vite/Context) • Vue.js • Tailwind CSS
Backend: Node.js • TypeScript • Python (FastAPI/Flask) • RESTful APIs
Data & Cloud: PostgreSQL • Redis • MongoDB • Firebase • Docker • CI/CD (GitHub Actions)
Business Tech: Stripe API (Complex Subscriptions) • OAuth 2.0 • Product Analytics

• 📍 Based in NYC (Green Card holder, no sponsorship required).
• 🏢 Former CEO & Founder: Ex-Drinkko & 4TIMP. I build with "Product Ownership" in mind.
• 🎓 Continuous Learner: Music conservatory background 🎸 → 40+ Technical Certifications.
• 🌍 Trilingual: English, French, Spanish (Fluent) — and I ship bilingual speech systems.
• 🛠 Workflow: Cursor & Claude Code power user (Agentic Dev workflow).

📧 info@mart-in.us
🔗 LinkedIn • Portfolio • Certifications

"The quick shall inherit the earth. I ship clean, performant code every week — measured, not assumed."