If you're building something ambitious in Backend Engineering, Computer Vision, AI Systems, or Camera Security — let's talk.
"I don't just write code — I architect systems that think, see, and scale."
I started with a simple question: what if a program could understand the world the way we do? That curiosity led me deep into the Python ecosystem — from building ironclad backends in Django, FastAPI, and Flask, to training object-detection models that watch the world through a camera lens — live, real-time, at the edge.
Today I sit at the intersection of disciplines most developers treat as entirely separate: Scalable Backend Engineering, Computer Vision / AI, and Physical Security Systems. I believe the best systems are the ones where these worlds talk to each other — where a trained YOLO model feeds a live API, where an ONVIF-connected camera streams frames over RTSP into a detection pipeline, where a Celery worker processes alerts asynchronously while Redis keeps state fast.
That's the kind of engineer I'm becoming — one sprint at a time.
| Domain | Technologies |
|---|---|
| Languages & Core | Python · JavaScript · C · ASP.NET · PHP · Bash |
| Backend & Web | Django · FastAPI · Flask · REST API · Token Auth · WebSockets |
| Task Queues & Caching | Celery · Redis · Pub/Sub · Rate Limiting · Background Workers |
| Containers & Orchestration | Docker · Docker Compose · Kubernetes · Pod Management · Service Mesh |
| Data & AI | NumPy · Pandas · PyTorch · Matrix Ops |
| Vision & ML | OpenCV · YOLO · ROI Tracking · Streamlit · VisionForge |
| Camera & Surveillance | ONVIF · RTSP/RTP · FFmpeg · XVR/NVR Integration · H.264/H.265 · WebRTC · PTZ Control · Live Stream Ingestion |
| Databases | PostgreSQL · SQLite · CRUD Pipelines |
| Concurrency | AsyncIO · Multiprocessing · Threading · GIL · venv |
| Document Processing | PyPDF · PyMuPDF · PDF Slicer 2.0 · Ingestion Pipelines |
| AI Paradigms | RAG Pipelines · AI Agents · Metaclasses · Dynamic Classes |
| Web Automation | Playwright · GhostFetch · Headless Scraping |
| Data Validation | Pydantic · Dataclasses · Type Annotations |
| Anthropic / Claude | Claude API · Claude Code · Prompt Engineering · Agentic Workflows · RAG with Claude |
| Auth & Validation | JWT · Token Auth · SQLAlchemy · Pydantic-AI · UV |
| Biometrics & Spatial | MediaPipe · PalmID · AeroPiano · Gesture Recognition |
| Security & Surveillance | ONVIF · RTSP/RTP · HLS · MJPEG · GStreamer · SSL/TLS · OAuth2 · OWASP |
| Edge AI & Deployment | ONNX · TensorRT · Raspberry Pi · On-device Inference |
| AI Security | DeepFace · InsightFace · LPR/ALPR · Anomaly Detection |
I think of software in systems, not scripts.
|
🟢 Backend Engineering Three frameworks, one philosophy — Django for batteries-included production, FastAPI for async-first, high-throughput APIs, and Flask when lightweight control is what the system needs. I pair every API with WebSockets for real-time channels, Celery for heavy background tasks, and Redis for caching and pub/sub. Not tools — a stack. 🟣 AI Agents & RAG The next frontier isn't just models — it's models that act. I've built terminal-based AI agents and RAG pipelines that make LLMs genuinely useful over private data. I've explored RAG hands-on using the Claude API — querying private corpora through a live Claude model, not just reading about it. 🔴 Machine Learning I've gone from zero to training, deploying, and evaluating my own YOLO models in a single month. The journey taught me that ML is 20% modeling and 80% data — cleaning it, annotating it, pipelining it. VisionForge was my answer: a full training toolkit built from scratch, with a Streamlit UI on top. |
🔵 Camera Security Systems From pulling live RTSP streams over ONVIF to integrating with XVR/NVR hardware — I work at the edge where physical security meets software intelligence. FFmpeg pipelines decode H.264/H.265 frames in real time, feeding directly into YOLO detection models. PTZ control, stream health monitoring, multi-camera orchestration, and device discovery over IP networks are first-class engineering problems in my stack. Sentrix-AI is the result. 🐳 Containers & Orchestration Shipping code is one thing. Shipping systems that stay up is another. I containerize pipelines with Docker, compose multi-service environments cleanly, and manage deployments with Kubernetes — pods, services, scaling, health checks. Infrastructure is not an afterthought; it's the last mile between an idea and a running product. ⚡ Concurrency The GIL is not an enemy — it's a puzzle. I've shipped multiprocessing solutions that bypass it entirely, async pipelines that never block, and threading models that stay safe. Performance isn't a feature; it's a discipline. 🤖 Claude & Anthropic I build with Claude, not just alongside it. From wiring the Anthropic API into RAG pipelines to using Claude Code as a terminal-first dev partner — I treat Claude as infrastructure. Prompt engineering, agentic chaining, private-data retrieval: these are first-class engineering concerns in my stack, not afterthoughts. |
Where computer vision meets physical security infrastructure.
Sentrix-AI is not a security wrapper — it's a ground-up intelligent surveillance system built to prove that open-source AI can replace expensive proprietary analytics engines.
|
Building multi-step agentic workflows beyond single-prompt RAG. LangGraph's stateful graphs are the next layer on top of the AI Agents & RAG pipelines I've already built. |
Going deeper into transformer architectures and fine-tuning pre-trained models. Running local LLMs via Ollama — private, fast, no API costs. |
Using the Anthropic API to wire Claude into real pipelines — RAG, agents, document Q&A. Using Claude Code for terminal-first agentic dev. Anthropic's ecosystem is core infrastructure in my stack. |
Adopting Pydantic for rigorous data validation and typed models. Using Playwright for headless automation in GhostFetch — my open-source scraping toolkit. Clean data in, clean data out. |
| Status | Domain | Topics |
|---|---|---|
 |
Computer Vision | YOLO · OpenCV · ROI · VisionForge · Sentrix-AI |
|
Camera Security Stack | ONVIF · RTSP/RTP · FFmpeg · XVR/NVR · PTZ · H.264/H.265 · Stream Ingestion |
|
Backend Engineering | Django · FastAPI · Flask · REST API · Token Auth · WebSockets |
|
Async & Task Queues | Celery · Redis · Pub/Sub · Background Workers · Rate Limiting |
|
Containers | Docker · Docker Compose · Multi-service Environments |
|
RAG & Agents | Ingestion Pipelines · Terminal AI Agent · Claude API RAG |
|
Document Processing | PyPDF · PyMuPDF · PDF Slicer 2.0 · Multiprocessing Pipelines |
|
Web Automation | GhostFetch · Playwright · Headless Scraping · Pydantic |
|
Anthropic Stack | Claude API · Claude Code · Prompt Engineering · Agentic Workflows |
 |
Kubernetes | Pod Management · Services · Scaling · Health Checks |
|
Transformers | Hugging Face · Fine-tuning · Embeddings |
|
LLM Orchestration | LangChain · LangGraph · Multi-Agent Systems |
|
Local AI Stack | Ollama · Private LLMs · On-device Inference |
|
Data Validation | Pydantic-AI · UV · Advanced Type Systems |
 |
Deep Learning | Neural Architectures · Backprop · CNNs · Attention |
|
MLOps | Model Versioning · CI/CD for ML · Monitoring in Production |
"Every line of code is a decision. Every system is an argument about what matters."
