AI Agentic Systems Engineer · Full-Stack Architect · Cloud & Automation Specialist

I design and ship production-grade AI products that create measurable business value.
My work spans agentic AI, LLM application engineering, retrieval-augmented generation, OCR and document intelligence, computer vision, full-stack platforms, cloud architecture, workflow automation, and operational analytics.

---

AI Systems I Build

🤖 Agentic AI & LLM Systems Tool-using language-model agents that coordinate APIs, databases, search, business rules, and human approvals. I focus on structured outputs, evaluation, observability, guardrails, and reliable fallbacks—not demos that fail silently.	📄 RAG & Document Intelligence OCR, retrieval-augmented generation, entity extraction, classification, validation, and review workflows for financial, legal, operational, and knowledge-management use cases.
🧠 Computer Vision & ML Platforms PyTorch and YOLO pipelines, real-time inference applications, ONNX deployment, telemetry, measurement tooling, model evaluation, and practical MLOps for desktop, edge, and cloud environments.	☁️ Full-Stack & Cloud Architecture React and Next.js interfaces, Python and Node.js services, PostgreSQL and Redis data layers, containerized delivery, infrastructure as code, CI/CD, and production monitoring on modern cloud stacks.

---

Selected Delivery Outcomes

Business outcome	Measured impact
LLM agent for support triage	96% intent-match accuracy
OCR-driven document workflow	60% faster processing
Automated validation pipelines	90% fewer operational errors
Executive and operations dashboards	Real-time decision visibility

---

Featured Engineering Work

Project	What it demonstrates	Core stack
TeamDoc — Real-Time Collaborative Docs	Collaborative document editing with CRDT synchronization, WebSocket persistence, JWT authentication, sharing rules, presence, and team workflows.	Django, Channels, Next.js, Tailwind CSS, Yjs
YOLO Inference Studio	A desktop computer-vision application with webcam and RTSP inference, PTZ camera control, SAHI slicing, history replay, annotation, CSV export, and performance analytics.	Python, PyQt5, Ultralytics YOLO, OpenCV, pandas
MASF-YOLO for Aerial Small-Object Detection	An unofficial PyTorch implementation of a YOLO11 architecture designed for tiny, densely packed objects in drone imagery, including custom feature aggregation and attention modules.	PyTorch, Ultralytics, OpenCV, Python
SAC GRU Distributed Training & Visualization	Soft Actor-Critic training for Simulink DLL environments with MLP and GRU policies, ONNX export, inference tooling, and desktop/web telemetry dashboards.	PyTorch, ONNX, PyQt5, Dash, Plotly

Explore all repositories →  ·  View client and product case studies →

---

Technical Stack

AI, Machine Learning & Computer Vision

Languages, Backend & APIs

Frontend & Product Interfaces

Data, Cloud & DevOps

---

GitHub Activity

3D Contribution Map

Contribution Stream

---

🧭 Engineering Principles

• Ship small, instrument early, and iterate from real usage.
• Use feature flags, staged rollouts, and reversible changes for AI features.
• Treat evaluation, traces, latency, cost, and failure modes as product requirements.
• Prefer explicit schemas, deterministic validation, and human approval for high-impact actions.
• An AI system that cannot explain or trace its decisions should not make irreversible production changes.

---

Let’s Build Something Valuable

Building an AI product, LLM agent, RAG platform, document-intelligence workflow, computer-vision system, cloud-native application, or operations automation program?