M. Tayyab Haider

M. Tayyab Haider

CS at LUMS | AI Researcher & Full-Stack Builder | Vocal Performer

Based in: Lahore, PakistanCurrently: Vocal Performer, IndependentExperience: 1 year

Hire me

Computer Science student at LUMS and 100% Merit Scholar with a deep focus on Computer Vision and AI-driven applications. I've built everything from wildlife detection pipelines with 98% accuracy to full-stack AI note-taking tools. Beyond the technical, I'm a seasoned vocal performer and leadership figure in the LUMS Music Society, bringing a unique blend of creative energy and engineering precision to every project.

Experience

Vocal Performer

IndependentfreelanceOct 2025 – Present

Summary

Performed at paid live events and private functions, managing client communication, scheduling, and on-site coordination.

What I did

  • Built a personal performance portfolio through consistent audience engagement and referrals.

Research Intern

LUMS Computer Vision & Graphics LabworkJun 2025 – Present

Summary

Animal Detection in Camera trap Images & Image Deraining

What I did

  • Developed a deep learning pipeline for wildlife detection in camera trap images using YOLO and U-Net, reducing false positives caused by rain artifacts.
  • Built and trained a U-Net model for rain detection and integrated it with a YOLO-based object detection pipeline, achieving 98% rain vs animal classification accuracy.
  • Designed a synthetic dataset of 10,000+ rain-degraded images and implemented a CNN-based kernel estimator to improve image recovery from rain deterioration.

Results

  • Achieved 98% rain vs animal classification accuracy.
YOLOU-NetCNNDeep LearningComputer Vision

Skills

Technical

PythonPython
Team Management
Leadership
Jupyter Notebook
PyTorchPyTorch
ReactReact
C++C++
FigmaFigma
MongoDBMongoDB
Microsoft Excel
Canva
VercelVercel
STATA
SQLSQL
LlamaIndex
Langchain
Pandas
RAG

Areas of expertise

Coordination and Logistics
Time Management
Problem Solving
Ownership
Communication

Projects

RAG Wikipedia Chatbot

personal
Built a retrieval-augmented generation chatbot that dynamically fetches and indexes Wikipedia articles using a ReAct agent, grounding every answer in retrieved facts. Implemented the full RAG pipeline: Wikipedia ingestion, document chunking, local HuggingFace embeddings (BAAI/bge-small-en-v1.5), vector indexing, and semantic similarity retrieval. Deployed session-isolated multi-user chat interface with model switching (LLaMA 3 / DeepSeek) via async Python and Chainlit's per-user session management.
PythonRAGLlamaIndexLangchainLLaMA 3DeepSeekChainlitHuggingFace

Camouflaged Object Detection (Wildlife Monitoring)

Deep Learning Studentacademic
Fine-tuned MegaDetector (MDv6) on COD10k dataset to improve detection on camouflaged wildlife. Evaluated model performance using Intersection over Union (IoU) and qualitative bounding-box visualizations to validate detection improvements.
Deep LearningMegaDetector (MDv6)COD10k dataset

LUMS Markaz – Campus E-Commerce Platform

HCI Studentacademic
Designed a campus-wide mobile e-commerce platform, addressing unmet student demand for secure on-campus transactions. Validated product decisions through cross-school user research, conducting 100+ surveys and 50+ interviews and app usability tests to refine UX around trust, discoverability, and usability.
Human Computer InteractionUX DesignUser Research

Urdu News Article Classification

Machine Learning Studentacademic
Scraped and curated 1,000+ Urdu news articles from Geo Urdu, Jang, and Express News across five categories: business, entertainment, sports, science-technology, and world. Cleaned, standardized, and vectorized the text data, then trained logistic regression, multinomial Naïve Bayes, and neural network models for multiclass classification. Achieved a peak test accuracy of 97.4% with the neural network; multinomial Naïve Bayes reached 96.4%, and logistic regression reached 95% test accuracy.
Machine LearningLogistic RegressionMultinomial Naïve BayesNeural NetworksWeb Scraping

AI Based Note Taking Web Application

Full-stack Developer & Product Visionarypersonal
Developed an all-in-one AI note-taking platform to solve the friction of switching between PDFs and external LLMs. Built a seamless workflow where users get real-time AI assistance—including text-to-speech, summarization, and automated quiz generation—directly within their note-taking environment.
ReactPythonMongoDBVercelAPI

Media

Education

B.Sc. Computer Science

Lahore University of Management SciencesComputer Science2022 – Jun 2026

Coursework

Introduction to Artificial IntelligenceMachine LearningDeep LearningSoftware EngineeringHuman Computer InteractionDatabasesCoding for Careers

Recognition

Awards

Merit Scholarship (100%)

The City School2020

Awarded for A levels

Governance & advisory

Director & Event Head | Electronics Category

LUMS Music Societyboard memberSep 2023 – Jun 2025

Led a 14-member design team and handled electronics-category operations for a 3-day LUMS music festival, coordinating participants, judges, volunteers, and logistics. Built a scalable design workflow and onboarding process that improved collaboration, enabled contributors of different skill levels to participate, and kept deliverables on schedule.

  • Led a 14-member design team and handled electronics-category operations for a 3-day LUMS music festival, coordinating participants, judges, volunteers, and logistics.
  • Built a scalable design workflow and onboarding process that improved collaboration, enabled contributors of different skill levels to participate, and kept deliverables on schedule.

Impact

  • Updated dates and added team leadership details for Music Society role.

What drives the work

Motivations

  • Delivering products to clients at speed
  • Building and iterating in fast-paced environments
Hire M.Reach out about a role, a contract or a conversation.For recruiters

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

More people on Kanvis

Browse more peopleCreate your own Kanvis