Machine Learning Researcher

Oct 2021 - Oct 2022

ChaLearn, U.S.A.

Front-End Engineer

Mar 2022 - April 2022

OpenML (openml.org)

Teaching Assistant

Oct 2021 - Mar 2022

Prof. Isabelle Guyon, Université Paris-Saclay

Software Development & IT Trainer

Jan 2019 - Aug 2020

Urraan, Abbottabad Pakistan (Urraan.pk)

Android Development Trainer

Jun 2016 - Aug 2016

TechValley, Abbottabad Pakistan

Machine Learning

Computer Vision

Meta-Learning

Deep Learning

Data Science

Fainess in ML

Challenge Organization

Project Management

Project Planning

Project Design

Product Development

Software Engineering

Software Development

Requirement Elicitation

Requirement Analysis

Mobile Applications Devlopment

IOS Development

Web Applications

Web Services

Innovation & Entrepreneurship

Startup Management

Database Design & Development

Research Writing

Project Proposal Writing

Technical Report Writing

Code versioning (Git/Github)

Github Pages

Docker

Linux

NeurIPS Checklist Assistant

An LLMs based checklist assistant for NeurIPS submissions

This study explores the use of Large Language Models (LLMs) as an assistant to help authors verify their submissions against the NeurIPS Paper Checklist. The goal is to assess whether LLMs can improve submission quality at NeurIPS. Participants receive feedback from an experimental LLM assistant to check compliance with NeurIPS submission standards. The LLM provides detailed feedback on the checklist responses to help authors refine their papers before submission. While the tool offers valuable guidance, it is meant to complement, not replace, the author's judgment and expertise.

Contributions: Machine Learning, Challenge Organization, Prompt Engineering, LLM integration, Data preparation

FAIR Universe

Unbiased Data Benchmark Ecosystem for Physics

The FAIR Universe project, funded by the US Department of Energy, is a collaboration between Lawrence Berkeley National Laboratory, Université Paris-Saclay, University of Washington, and ChaLearn. The initiative aims to create a large-scale AI platform for hosting scientific datasets, models, and machine learning competitions to advance discoveries in high energy physics and cosmology. The project focuses on reducing systematic uncertainties in High Energy Physics through a series of challenges. Key events include a toy challenge (October 2023), a Particle Physics hackathon (November 2023), and the HiggsML Uncertainty Pilot Competition (March 2024). A major challenge on uncertainties in fundamental science is set to launch at NeurIPS 2024. The project is ongoing and will conclude in 2025.

Contributions: Machine Learning, Challenge Organization, Software Development (front-end and back-end) to improve Codabench by adding more features and resolving current issues.

Stylized Meta-Album

Muti-domain computer vision meta-dataset

The Stylized Meta-Album (SMA) is a new image classification meta-dataset featuring 24 datasets (12 content and 12 stylized) to support research in out-of-distribution (OOD) generalization and related areas. SMA combines diverse subjects and styles, creating 4800 groups that offer extensive variability for rigorous studies. It introduces benchmarks for OOD generalization and group fairness, as well as unsupervised domain adaptation (UDA), showing the importance of group diversity in fairness and algorithm rankings, while also reducing error bars in benchmarking scenarios.

Contributions: Data preparation and processing, Data Analysis, Data Visualizations, Machine Learning, Website Development

Fairness in AutoML

Bias Mitigation in Machine Learning and AutoML
(Confidential : for internal use at Crédit Agricole)

A python library for bias detection and mitigation.

Contributions: Bias detection in structured data, Bias mitigation algorithms

Meta-Album

A meta-dataset for few-shot image classification

Machine learning for low-data regime is an area of interest in the research. Few-shot learning is one popular way to tackle this problem. The lack of good, challenging and computationally feasible datasets is a hurdle in smooth progress in few-shot learning. To remedy this, we created Meta-Album meta-dataset for few-shot learning, meta-learning and other purposes. It consists of 40 datasets from different domains. This research is supervised by Professor Isabelle Guyon (LISN,Université Paris-Saclay, ChaLearn) and is submitted to NeurIPS 2022.

Contributions: Machine Learning, Deep Learning, Meta-Learning, Few-shot learning, Data preparation and processing, Website Development

Deep Pollination

A series of ML challenges for classification of pollinating insects

Insects are very important for biodiversity, food chains, andpollination. It is of great importance to recognize insects, their habitats and to secure their natural environment. Machine learning, especially deeplearning techniques can be used to recognize and classify various insects. We introduce Deep-Pollination, a series of three machine learning challenges organized on Codalab for insects classification. A preprocessed version of the insects dataset (provided by MUSÉUM NATIONAL D’HISTOIRE NATURELLE) is used for these challenges which consistsof five classes and more than 200,000 images. This project is supervised by Professor Isabelle Guyon (LISN,Université Paris-Saclay, ChaLearn).

Contributions: Machine Learning, Deep Learning, Computer Vision, Challenge Organization

HCI Interface Design

Design of easy to use user interface for illiterate and semi-literate people

A study to design easy to use user interface for illiterate and semi-literate people to use mobile devices and computers easily. The study focuses on the use of images, animations, icons, video and background audio recordings and combinations of these to design user interfaces which enable these people to easily use computer platforms for job hunting, information gathering and for some other purposes.

Police CRMS

Malakand Levies Criminal Records Management System

A centralized and computerized system to replace the old files system and to equip the officials, office staff and check-posts staff to get updated data anywhere. HCI design principles are used in this project to make the transition from old system easy and to make it easier for semi-literate staff to use the new system.

Spotter Application

A software based solution for selling and buying parking spaces in Milan, Italy.

Contributions: Project Management, Design, Development and Deployment, IOS Development, Database design

Kardaan Application

A software based solution and established startup for hiring handymen services in Pakistan and Australia

Contributions: Project Management, Design, Development and Deployment, IOS Development, Database design, Web services, Startup development

Hello Demo Application

An application for music makers to share their work in their network and get their music signed

Contributions: IOS Development, Application deployment

Blood Community Application

An app based social services app to connect blood donors and receivers on one platform

Contributions: IOS Development, Database design, Web services

YÔGA Real Estate Solution

A platform for renting apartments, rooms, office spaces etc. in Paris and Shanghai

Contributions: Project Management, Web Development, IOS Development, Database design, Web services

Mailman Group

A website for Shanghai based Sports consultancy agency

Contributions: Project Management

Guess the Word Game

A mobile game for strengthening brain power using gamifying knowledge

Contributions: IOS Development

Stylized Meta-Album: Muti-domain computer vision meta-dataset

Submitted

Journal of Data-centric Machine Learning Research

RelevAI-Reviewer: A Benchmark on AI Reviewers for Survey Paper Relevance

Published

Conf´erence sur l’Apprentissage automatique (CAp) 2024

Meta-Album: Multi-domain Meta-Dataset for Few-Shot Image Classification

Published

Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks (2022)

Cross-Domain MetaDL: Any-Way Any-Shot Learning Competition with Novel Datasets from Practical Domains

Accepted

NeurIPS 2022 Competition Track (2022)

Lessons learned from the NeurIPS 2021 MetaDL challenge: Backbone fine-tuning without episodic meta-learning dominates for few-shot learning image classification

Published

PMLR - Proceedings of the NeurIPS 2021 Competition and Demonstration Track (2022)

MetaDL: Few Shot Learning Competition with Novel Datasets from Practical Domains

Accepted

NeurIPS 2021 Competition Track (2021)

Deep-Pollination: A series of ML challenges for multi-class insects classification

Accepted (Poster)

Junior Conference on Data Science and Engineering (2021)

Data Efficient Learning: A comparison of Transfer Learning and Meta-Learning

Accepted (Poster)

Junior Conference on Data Science and Engineering (2021)

Masters Excellence Scholarship

2020 - 2022

Université Paris-Saclay / Labex Digicosme

University Gold Medal

2017

University of Engineering and Technology Peshawar

Winner of 1 year Incubation for Startup KARDAAN

2019 - 2020

National Incubation Center, Islamabad Pakistan

8th position all over the country – Huawei ICT Skills Competition

2016

Arranged by Huawei

2nd position in province – KP Apps Challenge

2014

Arranged by Khyber Pakhtunkhwa (KP) Information Technology Board and World Bank

President, Academics Society

Sep 2016 - July 2017

University of Engineering and Technology Peshawar

Vice President, Academics Society

Sep 2015 - Sep 2016

University of Engineering and Technology Peshawar

Events Organizer

Sep 2015 - July 2017

University of Engineering and Technology Peshawar

Basketball Coordinator, Sports & Recreational Society

Sep 2014 - July 2017

University of Engineering and Technology Peshawar

English

Urdu

Pashto

French

Hiking

Nature

Space & Astronomy

Fitness & Bodybuilding

Technology

Prof. Isabelle Guyon

Professor, Masters Artificial IntelligenceUniversité Paris-Saclay, France
guyon@chalearn.org

Prof. Kim Gerdes

Professor, Masters Artificial Intelligence Université Paris-Saclay, France
kim.gerdes@universite-paris-saclay.fr

Prof. Sadaqat Jan

Professor, DEAN Faculty of Computing University of Engineering & Technology Mardan, Pakistan
sadaqat@uetmardan.edu.pk

Prof. Ibrar Ali Shah

Professor, Dept. Computer Software Engineering University of Engineering & Technology Mardan, Pakistan
ibrar@uetmardan.edu.pk

Email

ihsan2131@gmail.com

ihsan.ullah@chalearn.org