I build intelligent data pipelines and AI systems that solve real production problems — not prototypes.
My background is in mission-critical infrastructure: years operating 24/7 environments where systems can't fail. That foundation separates the way I build from most: I understand deployment, reliability, and observability before a model or pipeline ever reaches production.
Today I focus on data engineering with AI applied at every layer — ETL pipelines, LLM systems, RAG architectures, and autonomous agents. Every project in this profile processes real data, runs in production, and solves measurable problems.
AI-augmented engineering is my daily workflow — Claude Code, Gemini CLI, GitHub Copilot, Cursor, OpenAI Codex, and Antigravity as active engineering partners, not just assistants.
Currently: Running a helpdesk analytics platform in production processing 21,000+ tickets across enterprise contracts (KFC · IMC) — real-time ingestion via Pipefy and InvGate webhooks, 5-layer SLA normalization, predictive breach analysis, and semantic classification with local LLMs (Qwen2.5 14B).
Intelligent Data Pipelines High-performance ETL and analytics with FastAPI, DuckDB, Polars, and PostgreSQL. Real-time ingestion, incremental sync, and operational intelligence at scale.
LLM Systems & RAG Retrieval-Augmented Generation pipelines with semantic chunking, pgvector, reranking, and output quality evaluation. Built for production — not tutorials.
Autonomous Agents Multi-step agents with LangChain and LangGraph that handle real workflows: ticket triage, knowledge retrieval, incident response, and decision routing.
AI-Augmented Infrastructure MCP Servers, observability layers, and automation tooling that integrates LLMs into existing systems without breaking what already works.
AI & Data
Models & Inference
AI Tooling
Engineering
The three projects below represent production systems, not exercises.
data-analyzer Helpdesk BI platform processing enterprise support operations. ETL pipeline with incremental sync, MTTR calculation against business hours, 5-layer SLA normalization, and statistical projections. Stack: Python · FastAPI · DuckDB · React 19 · TanStack Query.
py-rag-engine Production RAG engine with semantic chunking, pgvector persistence, and relevance evaluation. Built to serve as the retrieval backbone for AI assistants in high-volume support environments. Stack: Python · LangChain · pgvector · SQLAlchemy · pytest.
base-imc-lite Enterprise knowledge management platform with AI chatbot (GPT-4), 3-level RBAC, brute-force protection, JWT with JTI blacklist, and audit trail for 1,000+ events. Stack: Next.js · TypeScript · OpenAI · bcryptjs · jose.
