Skip to content
View syedahmadbokhari's full-sized avatar
  • Bradford, UK
  • 16:19 (UTC -12:00)

Block or report syedahmadbokhari

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
syedahmadbokhari/README.md

Syed Muhammad Ahmad Bokhari — AI & Data Professional

First Class Honours · BSc Applied Artificial Intelligence (University of Bradford, 2026)
📍 Bradford, UK · UK Resident (Post-Study Work Visa) · Open to relocate

👋 About Me

Recent AI/Applied AI graduate specialising in Data Engineering, ML Systems, and Environmental AI.

I build production-grade data pipelines, computer vision systems, and AI-powered analytics tools. My final year project was deployed with the University of Bradford as a live CCTV monitoring system for real-time illegal dumping detection.

Currently seeking Data Analyst, Data Engineer, or ML Engineering roles in the UK (Bradford area, willing to relocate).

🛠️ Tech Stack

Languages: Python, SQL (Advanced), TypeScript
Data Engineering: Apache Airflow, dbt, DuckDB, PostgreSQL, AWS S3
ML/Analytics: XGBoost, scikit-learn, TensorFlow, PyTorch, Pandas, SHAP
Computer Vision: YOLOv8, MobileNetV3, ByteTrack, OpenCV
Backend & Deployment: FastAPI, Flask, SQLAlchemy, Redis, Streamlit, Docker, GitHub Actions
Databases: PostgreSQL, DuckDB, SQLite, pgvector
Certifications: dbt Associate, dbt Core

⭐ Flagship Project

AI for Environmental Monitoring & Urban Planning

BSc Applied AI Final Year Project — Deployed with University of Bradford

A real-time AI system that transforms passive CCTV infrastructure into an intelligent environmental monitoring tool. It automatically detects illegal littering and dumping events, tracks objects across frames, and alerts authorities — all without facial recognition (GDPR compliant, ALTAI ethical AI framework).

🔧 System Pipeline

CCTV Feed → Motion Detection → ByteTrack Tracking → MobileNetV3 Classifier → Alert → Dashboard

✨ Key Features

  • 🎯 MobileNetV3 waste classifier (PyTorch) — lightweight, optimised for real-time edge inference on CCTV hardware
  • 🔍 ByteTrack multi-object tracking with Kalman filtering for persistent object IDs across frames
  • 🚨 Stationary object logic — flags objects that remain in place beyond a time threshold as dumping events
  • 📡 Flask REST API with MJPEG video streaming and live Server-Sent Events (SSE) for real-time alerts
  • 🗺️ Streamlit dashboard with GIS map visualisation, runtime event logging, and accountability tracking
  • ✅ GDPR compliant — no facial recognition, full data minimisation, transparent audit trail
  • 🤝 ALTAI ethical AI framework applied throughout design and evaluation

📁 Project Structure

core/         ← Detection pipeline & classifier modules
dashboard/    ← Flask backend + Streamlit frontend
models/       ← Trained MobileNetV3 checkpoints
training/     ← Model training notebooks
data/         ← Raw & processed datasets
docs/         ← Design & evaluation reports
archive/      ← Historical prototypes (v0–v7)

🛠️ Full Tech Stack

Layer Technology
Language Python
Computer Vision OpenCV, YOLOv8
Deep Learning PyTorch, MobileNetV3, torchvision
Object Tracking ByteTrack (Kalman filtering)
Backend Flask (REST API, MJPEG, SSE)
Dashboard Streamlit + GIS visualisation
Deployment University of Bradford (live CCTV)

📊 Other Notable Projects

UK Crime Data Pipeline

End-to-end data platform ingesting 21k+ monthly UK police records
S3 → dbt → Airflow → FastAPI → Streamlit · JWT auth · Redis caching · RAG with pgvector · 20+ tests · CI/CD

Breast Cancer Survival Model

Production ML pipeline with SHAP explainability, model calibration, and survival analysis
scikit-learn · XGBoost · Pandas · Streamlit

Retail Data Platform

Cloud-native analytics with dimensional modelling, recommendation engine, and 51 unit tests
dbt · DuckDB · AWS S3 · Streamlit · pytest · moto

CLIP: Zero-Shot vs Linear Probe

Benchmarking zero-shot CLIP against linear probe fine-tuning across vision datasets
PyTorch · OpenAI CLIP · scikit-learn

Reinforcement Learning Game Agent

Policy gradient and deep Q-network implementations
PyTorch · OpenAI Gym

🏆 Engineering Highlights

  • Real-time computer vision pipeline deployed on live University of Bradford CCTV infrastructure
  • 8-task Airflow DAG with incremental loading, data validation, and alerting
  • Production REST API: async SQLAlchemy, cursor pagination, rate limiting, Redis caching
  • LLM-grounded report generation with semantic search via pgvector (Gemini AI)
  • dbt marts with schema tests, lineage tracking, and documentation
  • Multi-stage Docker builds with CI/CD via GitHub Actions
  • Geospatial analytics and crime hotspot mapping

📧 Let's Connect

Pinned Loading

  1. AI-for-Environmental-Monitoring-and-Urban-Planning AI-for-Environmental-Monitoring-and-Urban-Planning Public

    Real-time illegal littering & dumping detection using MOG2, ByteTrack (Kalman), YOLOv8 suppressor and MobileNetV3 — BSc Applied AI FYP, University of Bradford

    Python

  2. UK-Crime-Data-Pipeline UK-Crime-Data-Pipeline Public

    End-to-end UK crime pipeline: S3 ingestion, DuckDB, dbt, Airflow, FastAPI (JWT/RAG/Redis), React dashboard - deployed on Railway.

    Python 1

  3. uk-retail-data-platform uk-retail-data-platform Public

    Advanced SQL analytics: window functions, CTEs, cohort analysis, and business reporting queries across complex multi-table schemas.

    Jupyter Notebook 1

  4. breast-cancer-analysis breast-cancer-analysis Public

    ML classification system for breast cancer diagnosis: feature engineering, cross-validation, model comparison (SVM/RF/LR) with Streamlit UI.

    Jupyter Notebook 1

  5. clip-zero-shot-vs-linear-prob clip-zero-shot-vs-linear-prob Public

    Comparing zero-shot CLIP prompt engineering (single, universal, class-specific ensemble) against linear probe fine-tuning on image classification.

    Jupyter Notebook 1

  6. Reinforcement-Learning-Game-Agent Reinforcement-Learning-Game-Agent Public

    Reinforcement learning game agent trained with policy gradient methods and deep Q-networks using PyTorch.

    Python