Zuhaib Ahmed

AI Engineer / Data Scientist / Aspiring Full-Stack Developer

AI Engineer · Data Scientist · Aspiring Full-Stack Developer · Multi-Agent Systems Builder

AI Engineer, Data Scientist, and aspiring full-stack developer with a strong foundation in machine learning and statistical analysis. I build and deploy real AI systems end-to-end — from model training and backend architecture to containerized, production-ready deployment. My current focus is agentic AI: designing multi-agent LLM pipelines that plan, research, self-critique, and synthesize, rather than relying on single-prompt wrappers. Equal parts data scientist and systems builder in the making, I care as much about whether something ships cleanly as whether it works in a notebook.

Education

Bachelor of Science in Computer Software Engineering — Sukkur IBA University, Sukkur, Pakistan (Continue)

Projects

ResearchPilot — Multi-Agent Autonomous Research System

Python, GitHub Models, FastAPI, OpenAI SDK, SSE, Docker

Built a multi-agent AI reasoning system (Microsoft Agents League Hackathon, Reasoning Agents track) that autonomously decomposes research questions, retrieves sourced evidence, self-critiques coverage, and synthesizes structured reports with citations.

Architected a 4-agent pipeline — Planner, Researcher, Critic, Synthesizer — using the OpenAI Python SDK against the GitHub Models free inference endpoint.

Implemented a self-critique re-research loop: the Critic agent detects knowledge gaps and triggers additional research when confidence falls below threshold.

Real-time web frontend using FastAPI + Server-Sent Events (SSE) streaming live pipeline progress; deployed on Hugging Face Spaces via Docker.

Multi-Agent AI Course Creator — Distributed Multi-Agent System with Google ADK

Python, Google ADK, A2A Protocol, Google Cloud Run, FastAPI

Built a distributed multi-agent system using Google's Agent Development Kit (ADK) and Agent-to-Agent (A2A) protocol that autonomously researches, judges, and builds course content.

Architected 4 microservice agents — Orchestrator, Researcher, Judge, Content Builder — each running in its own container and communicating via A2A.

Implemented workflow orchestration using LoopAgent and SequentialAgent with RemoteA2aAgent for cross-service communication.

Deployed on Google Cloud Run with a web app frontend displaying real-time agent progress and final course output.

AgroVision — AI Plant Disease Prediction System

Python, TensorFlow, Streamlit, Docker

End-to-end AI web app predicting plant diseases from leaf images using a CNN trained on 54,000+ images across 38 disease classes; 88% validation accuracy, 98% training accuracy.

Interactive Streamlit interface for real-time image upload, preprocessing (224x224 resize, normalization), and instant prediction; containerized with Docker.

Canteen Token / Order Queue System — Web-based queue system for Sukkur IBA University

PHP, MySQL, HTML, CSS, JavaScript, XAMPP

Web-based canteen token and order queue system for Sukkur IBA University, replacing manual queuing.

Three-Tier Architecture (Presentation, Business Logic, Data) with role-based access for Customer, Staff, and Admin.

Automatic sequential token generation, FIFO queue with real-time tracking, order status lifecycle, live notifications.

Applied Singleton, Factory, and Observer design patterns; full SDA documentation (Use Case, Class, Sequence, Activity, State UML diagrams).

FitWise — AI-Powered Fitness Assistant

Python, AI, Data Analysis

Fitness app generating customized workout plans based on fitness level, goals, and equipment; AI-driven healthcare advice and a weight-tracking visualization interface.

Movie Discovery App — Search, favorites & persistence

React 19, React Router, Context API, LocalStorage

Movie discovery web app with real-time search, add/remove favorites, and persistence via LocalStorage + Context API.

ScriptFlow AI — Video Script Generator

Python, AI, API Integration

AI-powered video script generator producing platform-aware scripts with hooks, scenes, and CTAs for YouTube, TikTok, Instagram, and Ads in under 30 seconds.

Task Management App with GUI — Desktop productivity app

Python, Tkinter, OOP

Desktop task management app with add/update/delete/view features, built with OOP principles.

Skills

Programming Languages: Python, C++, Java, JavaScript

Agentic AI & LLMs: Multi-Agent Systems, Reasoning Pipelines, Prompt Engineering, OpenAI SDK, GitHub Models (GPT-4o/GPT-4o-mini), LLM Orchestration, Server-Sent Events (SSE)

Data Science & ML: Machine Learning, Statistical Data Analysis, TensorFlow, Keras, Matplotlib, Seaborn, Pandas, NumPy, Scikit-learn, Deep Learning (CNN), Model Training & Evaluation

Web & Backend: FastAPI, Uvicorn, React 19, React Router, Context API, Django, HTML5, CSS, Tailwind CSS, Streamlit

DevOps & Deployment: Docker, Docker Desktop, Containerization, Hugging Face Spaces, Image Building, Port Mapping, Container Deployment

Databases & Tools: SQL, Git, GitHub, Google Cloud, Google Gemini, API Integration, Microsoft Office

Concepts & Design: OOP, Data Structures, Tkinter, Figma (UI/UX), Prototyping, Wireframing, Color Theory

Awards

SEEF Sindh Government Scholarship — Sindh Education Endowment Fund (Government of Sindh) ()

Certifications

Google Cloud Gen AI Academy APAC Edition (Cohort 1) — Hack2Skill (Apr 2026)

Intro to Machine Learning — Kaggle (Apr 2026)

Claude 101 — Anthropic (Mar 2026)

Maximize Productivity With AI Tools — Google via Coursera (Jul 2025)

Introduction to SQL — DataCamp (Feb 2026)

Agents League – Reasoning Agents — Global AI Community (Jul 2026)

Build Small — Certificate of Participation — Hugging Face × Gradio (Jun 2026)

Hackathons

ResearchPilot — Microsoft Agents League Hackathon (Reasoning Agents Track): A 4-agent autonomous research system — Planner, Researcher, Critic, Synthesizer — built on GitHub Models (free GPT-4o access) with a real-time streaming frontend.

MediAssist — Qwen Cloud Hackathon 2026 (MemoryAgent Track): A conversational AI health assistant with persistent memory — it remembers a user’s name, allergies, health history, and past conversations across sessions, using a dual memory system that separates what’s shown on screen from what the agent actually retains. Includes full authentication (login/signup) so health data stays private, plus a chat history sidebar for browsing past conversations.

Smart Photo Analyzer — Build Small Hackathon 2026 (Hugging Face + Gradio) (Backyard AI Track): Upload any photo and get instant AI analysis in English or Urdu — general description, text/receipt reading, plant or food identification, or plain-language explanation of medical images — with one-click Urdu translation. Built to make vision AI accessible to Urdu-speaking communities across Pakistan who are largely left out of English-first AI tools.

Contact: zuhaibmahar234@gmail.com