Available for opportunities

Imran Mansha

AI/ML Engineer specializing in Agentic AI, RAG Applications, LLM fine-tuning, deep learning, and intelligent healthcare systems. Building production-grade AI solutions from research to deployment.

LinkedIn ↗ GitHub → Download CV

Roles & Internships

10+

Projects

88.1%

NSCT Percentile

20+

Students Mentored

Scroll

About Me

Who I Am

Core Proficiency

RAG Systems80%

Agentic AI85%

Prompt Engineering70%

LLM Fine-Tuning80%

Medical Imaging AI75%

LLMOps & Deployment65%

Hi, I'm Imran Mansha, an AI Engineer focused on building intelligent systems powered by Large Language Models, Retrieval-Augmented Generation (RAG), and Agentic AI.

My expertise spans Generative AI, LLM Engineering, Deep Learning, Medical AI, and AI System Deployment. I enjoy transforming research ideas into production-ready AI systems that are efficient, explainable, and impactful.

ProfileAI Engineer

Experience6+ Months

DomainHealthcare, Ecommerce, Sports & Education

EducationBS in Information Technology

LanguagesEnglish, Urdu

LocationLahore, Pakistan

ToolsPyTorch, Hugging Face, LangChain, LlamaIndex, ChromaDB, FAISS, FastAPI, Docker

InterestsTraveling, Photography, Teaching, Sports

Download CV LinkedIn ↗

Technical Stack

Skills & Tools

🤖 AI / ML

Machine Learning Deep Learning Transformers Generative AI LLM Engineering RAG Agentic AI LLM Fine-Tuning LoRA / QLoRA Prompt Engineering Explainable AI

⚙️ Frameworks

PyTorch Transformers Scikit-learn HuggingFace 🤗 LangChain LangGraph LlamaIndex FastAPI

🗄️ Databases

FAISS ChromaDB Qdrant Pinecone PostgreSQL MySQL SQLite

📊 Data Science

Pandas NumPy Matplotlib SQL Feature Engineering Data Preprocessing Model Evaluation Data Visualization

🛠️ Tools & Platforms

Git GitHub Linux Docker VS Code Jupyter Hugging Face Hub Kaggle Groq Ollama

💡 Other Skills

AI Research System Design AI Agents LLMOps Problem Solving

Work History

Experience

JULY 2026 — SEP 2026

Software Development Intern

TDCP - Govt. Of Punjab — Bahawalpur

Built a visitor intake system with a live dashboard for TDCP's tourism office.
Developing BWP AI Travel Agent, an AIpowered double-decker bus reservation and seat-booking platform.
Building a FastAPI backend so a single admin action safely confirms payment, booking, and seat status together, avoiding the inconsistent states a manual.

NOV 2025 — DEC 2025

AI/ML Engineering Intern

DevelopersHub Corporation · Remote

Developed News Topic Classifier using BERT with AdamW optimizer, 2 epochs, batch size 16. Achieved 94.6% accuracy and 94.6% weighted F1-score.
Gained hands-on experience in prompt engineering, model training, and evaluation on real-world medical datasets.

MAY 2025 — JUN 2025

Machine Learning Intern

Arch Technologies · Remote

Fine-tuned LLaMA 3 for medical Chain-of-Thought reasoning using PEFT (QLoRA) and Hugging Face.
Developed an Ethereum price forecasting model using ARIMA, applying time-series analysis and model evaluation techniques.

JAN 2025 — APR 2025

Community Education Volunteer

Government High School · Bahawalpur

Organized awareness campaigns to encourage parents in a rural community to prioritize education over child labor.
Contributed to reducing educational inequality in a disadvantaged area through community-driven initiatives.
Conducted workshops and school events to motivate students and foster interest in learning.

What I've Built

Projects

01 // JUN 2026

AI-Based ECG Image Digitization System

Converted scanned paper ECG images into digital time-series signals using OpenCV morphological gridline removal, FFT suppression, and column-wise argmax pixel scanning. Added rule-based rhythm classifier detecting 5 arrhythmia categories with RR-interval analysis. Deployed as Streamlit web app with CSV export. Aligned with PhysioNet Challenge 2024.

OpenCV NumPy SciPy Streamlit Signal Processing Arrhythmia Detection

GitHub Repository →

02 // MAY 2025

Fine-Tuning LLaMA 3 for Medical Chain-of-Thought Reasoning

Fine-tuned Meta-LLaMA 3 (3B) Instruct on the FreedomIntelligence medical-o1-reasoning-SFT dataset using QLoRA on Kaggle's free GPU tier. Applied SFT via Hugging Face's SFTTrainer for domain-specific CoT medical reasoning. Evaluated pre/post fine-tuning performance using ROUGE-L scoring and deployed the adapter to Hugging Face Hub.

LLaMA 3 QLoRA SFTTrainer ROUGE-L HuggingFace Kaggle GPU

GitHub Repository →

03 // MAR 2026

Alzheimer's Disease MRI Classification Using Transfer Learning & XAI

Built an Alzheimer's MRI classifier using MobileNetV2 transfer learning with partial fine-tuning (last 30 layers), achieving 73.7% test accuracy on the ADNI dataset. Handled class imbalance via sklearn class weighting and applied augmentation across 70/15/15 splits. Integrated both Grad-CAM and LIME for Explainable AI.

MobileNetV2 Transfer Learning Grad-CAM LIME TensorFlow ADNI Dataset

GitHub Repository →

04 // DEC 2025

News Topic Classifier Using BERT

Fine-tuned bert-base-uncased on AG News dataset (120K samples) to classify headlines into World, Sports, Business, and Sci/Tech categories. Used AdamW optimizer, 2 epochs, batch size 16. Achieved 94.6% accuracy and 94.6% weighted F1-score. Deployed live interactive demo via Gradio for real-time predictions.

BERT Transformers AdamW AG News Fine-Tuning Text Classification

GitHub Repository →

More projects available on GitHub

View All on GitHub →

Credentials

Certifications

🎓

Machine Learning Specialization

Stanford University

10 July 2025 · Verify ↗

☁️

Generative AI for Healthcare

Google Cloud

27 August 2025 · Verify ↗

🏆

National Skill Competency Test (NSCT)

Higher Education Commission Pakistan

05 April 2026 · 88.1 Percentile · Verify ↗

📊

Career Essentials in Data Analysis

Microsoft

28 July 2024 · Verify ↗

🔬

Understanding Research Methods

University of London

17 July 2025 · Verify ↗

🤝

Soft Skills Training

OEC & ICPMD · EU-Supported Program

17 October 2025 · Verify ↗

📝

How to Write a Research Paper

The University of Faisalabad

17 July 2025 · Verify ↗