Profile banner
Deep Gandhi

Deep Gandhi

Machine learning engineer focused on LLMs and fairness

deep1401
Calgary, Alberta
Joined April 2026
Favicon imagedeep1401.github.ioFavicon imagegithub.comFavicon imagetwitter.comFavicon imagelinkedin.comFavicon imagescholar.google.comFavicon imagedeep1401.github.io

Network

3.6K connections

Summary

Researcher focused on fairness and debiasing in NLP — Deep Gandhi's academic work centers on mitigating bias in language models and exploring privacy-preserving training approaches (e.g., federated learning). He has co-authored papers on debiasing via Distributionally Robust Optimization and on federated approaches for hate speech detection and sentiment analysis in Indic languages. arxiv+3
Machine learning engineer and open-source contributor — As a Machine Learning Engineer at Transformer Lab, he authors tutorials and platform features (dataset generation, LoRA training, AMD GPU support, Runpod integration) that make model training and evaluation more accessible for researchers and practitioners. transformerlab+2
Bridges academic research and engineering practice — His profile shows a trajectory from academic research (MSc, research fellow roles, publications) into practical engineering roles and internships, indicating experience shipping ML systems and applying research ideas in real-world tooling. github+2
Teaching and mentorship experience — Served as a graduate teaching assistant at the University of Alberta for courses such as Basics of Machine Learning and Ethics of Data Science, with responsibilities including labs, office hours, grading, and student support. github

Work

Education

Projects

Writing

RobustDebias: Debiasing Language Models using Distributionally Robust Optimization

January 1, 2026

Proposes RobustDebias, an approach that adapts Distributionally Robust Optimization (DRO) to debias language models during fine-tuning, mitigating bias amplification across multiple demographics with minimal performance impact.

Favicon imagearxiv.org

Transformer Lab Now Works with AMD GPUs

May 1, 2025

Detailed guide and post describing how Transformer Lab added AMD GPU support, covering ROCm installation, PyTorch ROCm builds, permission issues, and WSL challenges.

Favicon imagetransformerlab.ai

Generating Datasets and Training Models with Transformer Lab

February 1, 2025

Tutorial and walkthrough demonstrating how to generate custom datasets from raw text and fine-tune models using LoRA adapters inside Transformer Lab.

Favicon imagetransformerlab.ai

A Federated Approach for Hate Speech Detection

January 1, 2023

Paper presenting a federated learning approach for hate speech detection; published in the proceedings of EACL 2023.

Favicon imagedeep1401.github.io

IndicFed: A Federated Approach for Sentiment Analysis in Indic Languages

January 1, 2021

Explores federated learning for sentiment analysis across multiple Indic languages, demonstrating privacy-preserving federated approaches can outperform some server-trained baselines for low-resource languages.

Favicon imagedeep1401.github.io