Abdul Hadi

Abdul Hadi

AI Researcher & Developer | RAG Specialist | CS at LUMS

Based in: Lahore, PakistanMost recently: Teaching Assistant - Software Engineering, Lahore University of Management SciencesExperience: 1 year

Hire me

Computer Science student at LUMS specializing in AI and RAG systems. Proven track record in building high-performance LLM applications, from medical assistants to financial KPI extraction pipelines. Experienced in optimizing retrieval performance, fine-tuning models with LoRA, and scaling gameplay engagement. Dedicated to bridging the gap between complex research and production-ready AI tools.

Experience

Teaching Assistant - Software Engineering

Lahore University of Management SciencesLahore, Pakistanpart time2026 – May 2026

Summary

Supervised projects of 3-4 groups in a 300-student Android development course. Guided students on software design, debugging, and implementation best practices.

Results

  • class of 300 students
Android DevelopmentSoftware Design

Teaching Assistant - Master's-level Data Science

Lahore University of Management SciencesLahore, Pakistanpart timeSep 2025 – Dec 2025

Summary

Assisted with grading, labs, and student support for Python and machine learning assignments.

What I did

  • Mentored students on core data science libraries including TensorFlow and NumPy.

Results

  • class of 50 students
PythonMachine LearningTensorFlowNumPy

AI Intern

NETSOL TechnologiesLahore, PakistaninternshipJul 2025 – Aug 2025

Summary

Developed a RAG-based financial assistant using PyTorch, FAISS, and embeddings, improving retrieval performance by 35%. Designed a KPI extraction pipeline using Python, NLP, and Pandas, reducing manual analysis effort by 70%. Processed 300+ pages of financial reports using OCR, chunking, and embedding pipelines. Benchmarked embedding models (MiniLM, bge-small) and LLaMA-based models for latency, accuracy, and hallucination reduction. Integrated the assistant into internal dashboards, improving analyst productivity.

Results

  • Improved retrieval performance by 35%.
  • Reduced manual analysis effort by 70%.
PyTorchFAISSPythonNLPPandasOCRMiniLMbge-smallLLaMA

Skills

Technical

PythonPython
RAG (Retrieval-Augmented Generation)
PyTorchPyTorch
ChromaDB
NLP
Machine Learning
LoRA
LLM Fine-Tuning
Pandas
FAISS
C#C#
Unity
Groq
Tavily
Render
VercelVercel
Android Development
ReactReact
TypeScriptTypeScript
React hooksReact hooks

Projects

RAG Financial Chatbot

AI Internprofessional
Built a RAG pipeline with embeddings, similarity search, and LLM-based reranking, improving answer relevance by 30+ points and achieving 90% accuracy. Processed financial reports exceeding 300+ pages, extracting KPIs and financial metrics for structured analysis. Evaluated embedding models (MiniLM, bge-small) across latency, retrieval quality, and accuracy benchmarks. Built end-to-end workflows for KPIs, balance sheets, cash flow, and financial diagnostics.
PythonPyTorchFAISSReactTypeScript

Mini LLaMA-like Transformer

Developerpersonal
Implemented a LLaMA-inspired transformer with RoPE, multi-head attention, and RMSNorm, reducing perplexity by 28%. Built training pipelines with gradient accumulation and mixed precision for efficient experimentation. Benchmarked against Hugging Face models, optimizing latency and memory usage.
PyTorch

LoRA Fine-Tuning

Developerpersonal
Applied LoRA adapters to large language models, reducing trainable parameters by 75% and GPU usage by 40%. Fine-tuned models on financial text for KPI extraction without full retraining. Improved convergence speed while maintaining strong downstream performance.
PyTorch

E-commerce Website

personal
Built an e-commerce website using React and TypeScript for a software engineering course.
ReactTypeScript

Tic Tac Toe Game

personal
Developed a tic tac toe game using the React framework.
React

AI Paint Visualizer

personal
Created an AI paint visualizer with a frontend built in React and TypeScript.
ReactTypeScript

MedEpi: Medical RAG Assistant

Lead Developer (Senior Project)personal
Developed a hybrid Web-RAG pipeline for medical question answering under Dr. Asim Karim. Implemented controlled query expansion and source reliability scoring to reduce hallucinations in epidemiological data.
PythonChromaDBGroqTavilyRenderVercelGemini EmbeddingsCrossEncoder rerankingBrave Search APITrafilaturaLlama 3.3 70BReactTypeScript

Media

Education

BSc. Computer Science

Lahore University of Management SciencesComputer ScienceGPA 3 / 42022 – 2026

O/A Levels

Lahore Grammar School Johar Town2017 – 2022

Recognition

Awards

100% Scholarship

Lahore Grammar School Johar Town

100% scholarship awarded for academic excellence during O/A Levels.

Hire AbdulReach out about a role, a contract or a conversation.For recruiters

Abdul's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis