I build AI systems that actually work in production — not demos, not notebooks, not "proof of concepts that need six months of cleanup before they're real." I'm talking stateful multi-agent orchestration, zero-hallucination pipelines, and LLM-powered platforms running at scale for real users.

Currently shipping agentic infrastructure at Sproxil and building VibeSynk under my AI venture PhononAI — a unified social media intelligence layer that treats content creation like an engineering problem.

Open to high-impact remote AI Engineer roles globally. If you're building something ambitious with LLMs or agentic systems, let's talk.

• Agentic system design — LangGraph, MCP, ReAct patterns, multi-agent orchestration with deterministic guardrails
• RAG pipeline architecture — vector search, document grounding, hallucination-resistant retrieval on GCP & AWS
• LLM integration at scale — model-agnostic factory patterns (Gemini, Claude, GPT, Grok), Pydantic v2 structured outputs, function calling
• Production cloud backends — FastAPI, Cloud Run, Redis, PostgreSQL, full observability and CI/CD
• Prompt engineering — structured, testable, evaluated prompts that behave predictably under real-world conditions

Enterprise conversational AI agent — 100k+ monthly active users

Took a rigid WhatsApp menu bot and replaced it with a closed-loop agentic system built on LangGraph. 12 deterministic tool definitions. LLM Factory Pattern for real-time model switching across Gemini, Claude, ChatGPT, and Grok. Redis-backed session memory. <500ms PIN verification latency. 99.9% uptime. Zero hallucinations on product verification data — by design, not by luck.

LangGraph MCP Pydantic v2 Redis FastAPI WhatsApp Business API MySQL

Agentic social media management platform — TikTok, Instagram, X unified

A single intelligent layer that manages cross-platform content, automates scheduling, and surfaces audience intelligence — without the creator having to think about each platform separately. Multi-agent orchestration pipelines under the hood. Currently in platform API integration and app store submission phase.

LangGraph Multi-agent orchestration FastAPI React PostgreSQL GCP Docker

Enterprise plant disease intelligence — real-time agronomic diagnostics

Model-as-a-Service architecture decoupling a visual inference layer (EfficientNet-B0 via PyTorch) from a conversational LLM reasoning layer (Gemini 2.5-Flash). Live optical flow via WebRTC — continuous real-time detection, no static uploads. Universal ML adapter supporting PyTorch, TensorFlow/Keras, and Scikit-Learn. ModelRegistry tracking F1, P50/P95 latencies. Full developer CLI, JWT RS256 + SHA-256 dual-auth, pytest coverage.

PyTorch EfficientNet-B0 Gemini WebRTC FastAPI Next.js 14 PostgreSQL Supabase

AI-powered insurance assistant — GCP Cloud Run

RAG-based insurance policy assistant with Vertex AI LLM inference, OAuth2 auth, and end-to-end observability via Cloud Monitoring. Reduced API latency by 20% through optimized query execution. Migrated CI/CD from GCP to AWS for multi-cloud flexibility.

Vertex AI RAG FastAPI Cloud Run OAuth2 Docker AWS

I've been going deep on Bitcoin protocol and Lightning Network development — specifically how to think about trustless, programmable value transfer as an infrastructure layer. It's the kind of low-level systems problem that I find genuinely fascinating after years of building at the application layer.

Early days. Intentional direction.

• 🤖 AI Agents & Agentic AI Systems Development · Meta AI Developer Academy (RAIN × Meta) — Dec 2025
• ☁️ AWS AI & ML Scholars — 2025 Challenge · AWS × Udacity — 2025
• 🚀 Google AI Startup School — Hands-on Labs · Google — Apr–Jun 2025
• 🎓 ALX Software Engineering — Back-end Specialization · ALX / Holberton Inc. — Nov 2023
• Also: Introducing Generative AI with AWS (Udacity/Accenture), AWS Educate ML Foundations, ICP TypeScript Azle (ICP Hub)

I'm genuinely interested in collaborating on agentic AI systems, production LLM infrastructure, and anything that pushes what software can do autonomously. If that sounds like you — reach out.