z
zahoor_develope

Zahoor Ahmad

@zahoor_develope

PRO

Pakistan
Engels, Urdu, Spaans
Sommige informatie wordt in het Engels weergegeven.
Over mij
I build AI systems that run in production, not notebooks that run once. My focus is computer vision and applied machine learning: custom YOLOv8 detection models, OCR pipelines, and the web apps around them. My last detection system hit 86.9% mAP at 43.7 FPS on live video and shipped with a working dashboard, not a folder of scripts. Full stack: Python, PyTorch, OpenCV, FastAPI on the model side. Next.js, React, Supabase, PostgreSQL on the product side. That means I deliver the whole thing. Same-day replies. I tell you upfront if your accuracy target isn't realistic.... Lees meer

Skills

z
zahoor_develope
Zahoor Ahmad
offline • 
Gemiddelde reactietijd: 1 uur

Bekijk mijn diensten

Full stack webapplicaties
I will build a custom yolov8 object detection model
VanafUS$ 60 / projectMeer informatie
Websites op maat
I will build a real estate website with CRM and lead capture
VanafUS$ 155 / projectMeer informatie

Portfolio

Werkervaring

Fiverr

Python AI Engineer | Computer Vision, LLM Agents + Full-Stack

Fiverr • Freelance

Nov 2025 - Present • 9 mos

You need a Python AI system that classifies, detects, predicts, or automates something real — and the usual freelancer hands you a Jupyter notebook and disappears. Most AI freelancers show up with a Kaggle notebook, fine-tune a model once, and vanish when the model drifts, the API bill blows up, or inference is too slow. Then you still have to find someone else to build the app around it. I do both: production Python AI, and the full-stack product it lives inside. COMPUTER VISION THAT RUNS IN REAL TIME - YOLOv8 object detection, OpenCV pipelines, EasyOCR text extraction - Shipped: motorcycle helmet violation detection — 86.9% mAP@0.50, 89.3% precision, 43.7 FPS on live video - Custom dataset prep, annotation, training, and tuning for your specific footage AI AGENTS THAT DO WORK, NOT JUST CHAT - LangChain, LlamaIndex, custom tool use - Agents that browse, write to your database, call APIs, and escalate to humans - RAG over your docs, Notion, websites, and internal wikis PRODUCTION ML, NOT NOTEBOOK CODE - scikit-learn, TensorFlow, PyTorch for classical ML and deep learning - End-to-end pipelines: data, training, evaluation, deployment, monitoring FULL-STACK DELIVERY — THE MODEL PLUS THE PRODUCT - Next.js, TypeScript, React front ends with real dashboards, auth, and role-based access - Node.js, FastAPI, Supabase, PostgreSQL back ends with row-level security - n8n automation to wire your AI into the tools your team already uses - Shipped: a multi-role school management platform now running as a live pilot, and a real estate CRM in production MY STACK AI + ML: Python 3.11+, PyTorch, TensorFlow, scikit-learn, NumPy, pandas, Keras Computer Vision: YOLOv8, OpenCV, EasyOCR, custom dataset training LLM: OpenAI (GPT-4o, Assistants, fine-tuning), Anthropic Claude, Google Gemini, prompt engineering, evals RAG + Vector DB: pgvector, ChromaDB, Pinecone, Qdrant, LangChain, LlamaIndex, embedding strategy Full-Stack: Next.js, TypeScript, React, Node.js, FastAPI, Supabase.