████████╗██╗  ██╗ ██████╗ ███╗   ███╗ █████╗ ███████╗
╚══██╔══╝██║  ██║██╔═══██╗████╗ ████║██╔══██╗██╔════╝
   ██║   ███████║██║   ██║██╔████╔██║███████║███████╗
   ██║   ██╔══██║██║   ██║██║╚██╔╝██║██╔══██║╚════██║
   ██║   ██║  ██║╚██████╔╝██║ ╚═╝ ██║██║  ██║███████║
   ╚═╝   ╚═╝  ╚═╝ ╚═════╝ ╚═╝     ╚═╝╚═╝  ╚═╝╚══════╝

Self-taught. Still learning. Building systems that think.

Torture your data until it confesses.

Building Pixel — a multi-agent system that reasons, answers, and acts on behalf of users. Ships VisionFlow — production computer vision for retail analytics (YOLO/RF-DETR, ByteTrack, active learning).

Production-grade CV system. Dual detector mode (YOLO11n / RF-DETR swappable via config flag), real-time multi-object tracking with ByteTrack, active learning loop that captures low-confidence frames for retraining. Runs fully local, no cloud dependency. Frontend dashboard in Next.js, inference server in FastAPI.

Input → Detection (YOLO / RF-DETR) → ByteTrack → Analytics → Dashboard
                                                          ↓
                                              Active Learning Loop

Stack: PyTorch · YOLO · FastAPI · Docker · MLflow · DVC · Next.js

An agentic system that reasons, queries, and acts for users. Designed for complex multi-step tasks with tool use, memory, and chain-of-thought. Being built solo.

Stack: Python · LangChain · Claude API · FastAPI

LangChain + Google Generative AI for multi-document RAG. Ask questions across PDFs, get grounded answers.

Streamlit app using Gemini as backend — drop a YouTube URL, get a script.

3 notebooks: auto-apply to jobs, scrape applied job details, generate ATS reports.

2023 ─── Jupyter notebooks, internships, classic ML
2024 ─── MLOps, production pipelines, LLMs
2025 ─── Computer vision systems, Docker deployments
2026 ─── Multi-agent systems, agent orchestration

45+ repos. Started with .fit() and .predict(). Now shipping containerized systems with active learning loops.

Actually use daily:

ML/DL:     Python · PyTorch · YOLO · HuggingFace · Scikit-learn
LLMs:      LangChain · Claude API · Gemini · GPT-4
Backend:   FastAPI · Streamlit · PostgreSQL · Redis
Infra:     Docker · Docker Compose · GitHub Actions
MLOps:     MLflow · DVC · Prometheus · Grafana
Frontend:  TypeScript · Next.js

Some things I've touched. Some I've shipped. Some I'm still learning.

Torture your data until it confesses.

Build things that work. Measure what matters. Prefer simple over clever.

• LinkedIn: linkedin.com/in/thomasantony666
• LeetCode: leetcode.com/7H0M45_4N70NY

Building. Breaking. Learning. Shipping.