📄 RESUME / CV — IMAD UD DIN
Download
AI Engineer · Pakistan 🇵🇰

IMAD
UD DIN

AI Engineer
Generative AI · Agentic AI · Deep Learning · ML

AI Engineer specializing in Generative AI, Agentic AI, and RAG-based systems. I design and build multi-agent pipelines, chatbots, and knowledge assistants using LangChain, LangGraph, CrewAI, AutoGen, and MCP — grounding LLM responses in real retrieved data, enabling autonomous tool use, and reducing hallucinations through careful prompt engineering. Backed by a strong Deep Learning foundation — CNNs, RNNs, transfer learning, and model optimization — applied to production-grade systems across NLP, computer vision, and medical imaging.

3Internships
12+Projects
1+Yrs Exp.
15+Tools
Technical Expertise

Skills & Stack

A comprehensive AI & data toolkit built through internships, projects, and relentless self‑learning.

🐍

Languages & Frameworks

PythonRPHP FlaskFastAPI HTML/CSS
🧠

Deep Learning

PyTorchTensorFlowKeras CNNsRNNs / LSTMGRU TransformersYoloV9VGG19
🤖

Generative AI & Agents

LangChainLangGraphMCP RAGLLMsPrompt Eng. Vector DBsAgentic AI OpenAI Agents SDKPydantic AI
⚙️

ML Techniques

Supervised LearningUnsupervised Learning Hyperparameter TuningTransfer Learning Ensemble MethodsPCA Scikit-learnPandasNumPy
🗄️

Databases & SQL

MySQLPostgreSQL SQL QueriesDatabase Design Schema ManagementCRUD Operations
🔧

Data Engineering

Data PreprocessingEDA Data CleaningData Transformation Feature EngineeringETL/ELT Pipelines Data AugmentationMatplotlib
Core Proficiency
Python & ML Libraries
Deep Learning (CNNs / RNNs)
LangChain & RAG Pipelines
Agentic AI & MCP
Computer Vision (YOLO / OpenCV)
Data Analysis & EDA
MySQL / PostgreSQL
ETL / Data Pipelines
Verified Credentials

Certificates

Credentials earned from Anthropic, reflecting hands-on fluency with Claude and agentic AI tooling.

✓
ANTHROPIC

Claude Code 101

Certificate of completion covering Claude Code — Anthropic's agentic coding tool for delegating development tasks via the command line and IDE integrations.

View Certificate →
✓
ANTHROPIC

AI Fluency: Framework & Foundations

Certificate of completion on AI fluency foundations — developed in partnership with University College Cork, Ringling College of Art + Design, and the Higher Education Authority.

View Certificate →
Work History

Experience

3 internships across government, private tech & international remote — building real AI systems.

ATS — United States
Remote · 2025
AI Engineering Intern
April 2025 — June 2025

Worked remotely for a US‑based company on advanced ML model optimization and deep learning pipelines. Focused on improving model accuracy and efficiency for regression and classification tasks.

  • Extensive hyperparameter tuning (layers, neurons, activations, learning rates) to optimize ANN performance
  • Applied PCA for dimensionality reduction — improved model generalization on high‑dimensional data
  • Transfer learning with ResNet, DenseNet, EfficientNet, YoloV9 on limited datasets
  • Implemented LSTM & GRU variants to capture long‑term dependencies & reduce vanishing gradients
  • Advanced data augmentation (rotation, flipping, brightness, noise injection) to prevent overfitting
PyTorchTensorFlowPCATransfer LearningLSTM/GRUYoloV9
Comrex Pvt Ltd — Islamabad
On‑site · 2024
AI Engineering Intern
July 2024 — September 2024

Built production‑ready Generative AI & Agentic AI systems. Designed scalable LLM pipelines and reusable AI components for real‑world deployment.

  • Built RAG pipelines using LangChain + vector databases for knowledge‑augmented LLM responses
  • Developed multi‑step reasoning workflows with LangChain agents & chains for task automation
  • Built Agentic AI systems enabling LLMs to autonomously use external tools & APIs
  • Applied prompt engineering & tuning to reduce hallucinations and improve output accuracy
  • Integrated GenAI into chatbots, Q&A systems & knowledge assistants
LangChainRAGLLMsAgentic AIPrompt Eng.Vector DBs
Ministry of IT & Telecom — Islamabad
Government · 2023
AI Intern
August 2023 — October 2023

First professional AI experience at a national government institution. Data analysis, visualization, and ML tasks within Pakistan's digital transformation initiatives.

  • Data preprocessing & EDA on structured datasets to extract actionable business insights
  • Built data visualizations using Python (Pandas, NumPy, Matplotlib) to identify trends & metrics
  • Applied foundational ML techniques for prediction and classification tasks
  • Contributed to research on AI applications in digital transformation & telecom
PythonPandasNumPyMatplotlibEDAScikit-learn
Selected Work

All Projects

12 real‑world projects spanning medical AI, computer vision, NLP, agentic AI, data engineering, AgriTech, backend APIs & full‑stack systems.

01 / 11Medical AI · Vision

Spine Disease Detection via X‑Ray

Automated spine disease classification using deep learning on X‑ray images — high accuracy on limited labelled DICOM medical data.

0.90 Validation AUC4,000+ X-ray images
  • YoloV9 object detection + CNN for accurate disease localization & classification
  • Grad-CAM++ for visual explainability — highlights model decision regions
  • Transfer learning for strong performance on limited medical imaging data
  • DICOM medical images handled via Pydicom preprocessing pipeline
PyTorchYoloV9Grad-CAM++OpenCVPydicomCNNs
02 / 11Medical AI · Vision

Brain Tumor Detection via MRI

Deep learning system for binary brain tumor classification from MRI scans using VGG19 transfer learning — tumor vs. no‑tumor with high precision.

VGG19 + ImageNet pretrainBinary classification
  • VGG19 pre‑trained on ImageNet, fine‑tuned for MRI brain tumor detection
  • Full preprocessing pipeline: resizing, normalization, one‑hot encoding
  • Training accuracy & loss visualization for model interpretability
  • Supports custom image inference — predict on new unseen MRI scans
TensorFlowKerasVGG19OpenCVNumPyScikit-learn
03 / 11ML · AgriTech

Vaccine Combo Prediction — Poultry

ML system predicting optimal vaccine combos for chicks — tackling missing data, high‑cardinality features & real agricultural complexity.

Stacking + boosting ensemblesFull EDA → Deploy
  • Ensemble methods (stacking & boosting) for robust prediction accuracy
  • Model‑based imputation to handle missing agricultural data records
  • Target encoding for high‑cardinality categorical variables
  • PCA for dimensionality reduction and improved generalization
Scikit-learnEnsemblesPCATarget EncodingPandas
04 / 11Agentic AI · Voice

Voice‑Based Personal Assistant

Intelligent voice‑activated assistant powered by Agentic AI — processes speech, reasons via LLM agents & executes real tasks through tools.

Whisper STT + LangChainEnd-to-end voice pipeline
  • OpenAI Whisper for accurate real‑time speech‑to‑text transcription
  • LangChain agent with tool‑calling for search, Q&A & task execution
  • Autonomous multi‑step reasoning for complex user queries
  • Extensible architecture — easily plug in new tools & capabilities
LangChainWhisperLLMAgentic AIPython
05 / 11ML · Computer Vision

Human Activity Recognition Model

ML model that classifies human activities from sensor data — deployed as a web application for real‑time inference with video support.

Trained .pkl modelWeb app deployed
  • Trained and serialized activity classification model (.pkl) from sensor data features
  • Feature extraction pipeline with engineered CSV features from raw activity data
  • Built Flask web app (app.py) for real‑time activity inference from new inputs
  • Supports video input for live human activity detection and classification
Scikit-learnPythonPandasFlaskJupyter
06 / 11Data Analysis · Telecom

Big Data Analysis for Telecommunications

Comprehensive big data analytics project for the telecom sector — full ETL pipeline, EDA, trend discovery & predictive analytics on large‑scale telco datasets.

Large-scale telco datasetETL + web dashboard
  • End‑to‑end data pipeline: ingestion, cleaning, transformation on telecom datasets
  • Deep EDA & visualization to uncover usage trends and customer patterns
  • Built Flask web interface (app.py) for interactive data exploration & reporting
  • Full research report documenting methodology, findings & business recommendations
PythonPandasNumPyMatplotlibFlaskJupyter
07 / 11Data Engineering · R

Data Cleaning & Preprocessing — Churn Dataset

End‑to‑end data cleaning and preprocessing pipeline using both R and Python on a real customer churn dataset — preparing raw data for ML model training.

R + Python dual pipelineCustomer churn dataset
  • R & Python dual‑language pipeline for data cleaning and transformation
  • Handled missing values, outliers, duplicates and data type inconsistencies
  • Feature engineering — created new variables and applied transformations
  • Produced clean CSV output (updated_churn_data.csv) ready for ML modelling
  • Full project report (PDF) documenting methodology and findings
RPythonPandasNumPyJupyter
08 / 11Full Stack · MySQL

Pharmacy Management System

Full‑featured web‑based pharmacy management system built with PHP & MySQL — handles medicines, customers, suppliers, invoices, stock & purchase reports.

PHP + MySQL backendFull CRUD system
  • MySQL relational database (pharmacy.sql) — full schema for medicines, customers, suppliers
  • Complete CRUD operations: add/manage medicines, customers, suppliers, purchases
  • Invoice generation (new_invoice.php) and sales & purchase reporting
  • Stock management — track medicine inventory levels in real time
  • Secure user authentication: login, logout, change password, profile management
PHPMySQLHTML/CSSSQL
09 / 11Agentic AI · RAG

Multi-Agent Research System

A fully agentic research pipeline: search, retrieve, write, and self-critique — built from scratch with LangChain, LangGraph & LCEL, with every tool call logged and verified.

4-agent pipelineTool-call verified
  • Search Agent retrieves live information via the Tavily API
  • Reader Agent autonomously scrapes sources — PDF-aware, auto-follows arXiv papers to full text, with fallback URLs on scrape failure
  • Writer & Critic chains built with LCEL synthesize a grounded report and self-score it
  • Full tool-call tracing verifies agents ground answers in retrieved data, not memory
LangChainLangGraphRAGAgentic AIGeminiStreamlit
10 / 12Backend · FastAPI

Patient Management API

A RESTful patient management API built with FastAPI — demonstrates clean endpoint design, request validation, and full CRUD on patient records.

Full CRUD + Swagger docsPydantic validated
  • Complete CRUD endpoints — create, view, update (partial), delete, and sort patients by age
  • Pydantic models for request validation, optional fields & Literal type checks
  • Partial updates via model_dump(exclude_unset=True) — only changed fields are written
  • Auto‑generated Swagger UI & ReDoc interactive API documentation
FastAPIPydanticPythonUvicornREST API
11 / 12Agentic AI · Discovery

Pakistan AI Company Discovery Agent

An agentic pipeline that discovers software houses & tech companies across Pakistani cities that genuinely work with AI — built to surface real leads for AI Engineer roles, internships & freelance work.

Deterministic-first, LLM only when ambiguousSQLite caching layer
  • Discovery + Classification agents orchestrate city search, scraping & AI‑domain tagging
  • Deterministic keyword classification first — a cached local LLM (Ollama + Gemma 3 4B) is only called on genuinely ambiguous pages, minimizing cost
  • SQLite caching for scraped domains & LLM responses — avoids re‑fetching or re‑classifying the same company twice
  • Streamlit UI — pick a city and AI domains, get back company name, city, AI domains, website & LinkedIn
Agentic AILangChain-style AgentsBeautifulSoupSQLiteOllamaStreamlit
12 / 12Agentic AI · Speech

AI Video Assistant

Turns any YouTube video or local audio/video file into a searchable transcript and chatbot — ask questions about the content and get answers grounded in what was actually said.

YouTube + local file supportChat over transcript
  • Ingests YouTube links or local audio/video files and produces a full transcript
  • Transcript is made searchable — quickly locate where a topic is discussed in the video
  • Chatbot interface answers natural‑language questions grounded in the video's actual content
  • Modular Utils pipeline separating ingestion, transcription & chat logic, with test coverage
PythonSpeech-to-TextAgentic AITranscript SearchChatbot
Academic Background

Education

Formal CS education running parallel with 2+ years of real‑world AI engineering experience.

🎓
Bachelor of Science — Computer Science
Pak‑Austria Institute of Applied Sciences & Technology
Haripur, Khyber Pakhtunkhwa, Pakistan  ·  Expected July 2026

BS Computer Science with a strong focus on Generative AI, Agentic AI, Deep Learning & Data Engineering. Alongside studies, completed 3 internships and 12+ independent projects spanning Generative AI (LLMs, RAG pipelines, prompt engineering), Agentic AI (LangChain, OpenAI Agents SDK, Pydantic AI), Deep Learning (ANNs, CNNs, RNNs), Backend APIs (FastAPI, Pydantic), and Full‑Stack Development (PHP, MySQL, Flask). Goal: full‑stack AI Engineer designing and deploying intelligent, autonomous systems at scale.

Let's Connect

Open to AI
Opportunities

Actively looking for internships, freelance AI projects & full‑time AI Engineering roles. ML pipelines, GenAI apps, Agentic systems — ready to contribute & deliver results. Remote and on‑site both welcome.

profile.json
// Imad Ud Din — AI Engineer
{
  "name": "Imad Ud Din",
  "title": "AI Engineer",
  "focus": "GenAI · Agentic AI · Deep Learning · Data Eng.",
  "github": "github.com/imaduddin005",
  "education": "BS CS · 2026",
  "internships": 3,
  "projects": 12,
  "databases": ["MySQL", "PostgreSQL"],
  "specializations": [
    "Deep Learning",
    "Generative AI & RAG",
    "Agentic AI Systems",
    "Data Engineering & ETL"
  ],
  "status": "open_to_work"
}