Skip to content
Khalid Saifullah
GitHubXEmail

Portfolio

Research

  • Scalable Pretraining of Retrieval Models. Khalid Saifullah, John Kirchenbauer, Rifaa Qadri, Abhimanyu Hans, Neel Jain, Yuxin Wen, Siddharth Singh, Brian R. Bartoldson, Bhavya Kailkhura, Abhinav Bhatele, Tom Goldstein. ICLR SCI-FM, 2025.

  • Bring Your Own Data! Self-Supervised Evaluation for Large Language Models. Khalid Saifullah*, Neel Jain*, Yuxin Wen, John Kirchenbauer, Manli Shu, Aniruddha Saha, Micah Goldblum, Jonas Geiping, Tom Goldstein. COLM, 2024.

  • CinePile: A Long Video Question Answering Dataset and Benchmarks. Ruchit Rawal, Khalid Saifullah, Ronen Basri, David Jacobs, Gowthami Somepalli, Tom Goldstein. CVPR SynData4CV, 2024. Best Paper.

  • Dynamic Guardian Models: Realtime Content Moderation With User-Defined Policies. Monte Hoover, Vatsal Baherwani, Neel Jain, Khalid Saifullah, Joseph James Vincent, Chirag Jain, Melissa Kazemi Rad, C. Bayan Bruss, Ashwinee Panda, Tom Goldstein. ICML MoFA, 2025.

  • Seeing in Words: Learning to Classify through Language Bottlenecks. Khalid Saifullah, Yuxin Wen, Jonas Geiping, Micah Goldblum, Tom Goldstein. ICLR Tiny Papers, 2023.

  • LiveBench: A Challenging, Contamination-Free LLM Benchmark. Colin White*, Samuel Dooley*, Manley Roberts*, Arka Pal*, Benjamin Feuer, Siddhartha Jain, Ravid Shwartz-Ziv, Neel Jain, Khalid Saifullah, Siddartha Naidu, Chinmay Hegde, Yann LeCun, Tom Goldstein, Willie Neiswanger, Micah Goldblum. ICLR, 2025. Spotlight.

  • On the Reliability of Watermarks for Large Language Models. John Kirchenbauer*, Jonas Geiping*, Yuxin Wen, Manli Shu, Khalid Saifullah, Kezhi Kong, Kasun Fernando, Aniruddha Saha, Micah Goldblum, Tom Goldstein. ICLR, 2024.

See the complete list on Google Scholar.

Industry Experience

  • Research Intern, Microsoft AI. Summer 2026.
  • Research Scientist Intern, Adobe Research. 2025.
  • Research Scientist Intern, Adobe Research. 2023.

Talks & Workshops

Awards & Community Contributions

Projects

  • DALL·E mini: A JAX/FLAX implementation of OpenAI’s DALL·E but ~30 times smaller and trained on a much smaller dataset to meet resource constraints. (source code)

  • GPT-2 Bengali: First implementation of GPT-2 from scratch for Bengali. Trained with the Bengali portion of multilingual-c4 dataset. (source code)

  • Bengali Lyricist: A deep generative model that generates Bengali song lyrics. The model is finetuned on the song lyrics dataset from kaggle. (source code)

  • BERTify: A python module that helps with extracting BERT embeddings for large Benglai/English dataset. Specially optimized for users with limited computational access (e.g. free colab/kaggle GPUs) to tackle CUDA OOMs. (source code)

  • DietNeRF: First coding implementation of Putting NeRF on a Diet paper that requires much fewer image samples than original NeRF model to perform 3D reconstruction. (source code)

  • Misogynistic Meme classifier: As part of AllenNLP Hacks 2021, trained a multimodal model for misogynous meme identification using pretrained CLIP as feature extractor. (source code)

  • Path Finder: A web app that helps to visualize the typical graph searching algorithm such as Dijkstra, A*, etc. (source code)

  • 8-Puzzle Solver: The 8-puzzle game and also lets the user get it solved by the A.I (A* heuristics based), made as part of the AI course project. (source code)

  • BOOKy: A website for book lovers to lend and borrow books independently, made with Django as part of System Analysis and Design course project.

  • One Step: A website that helps user to create a cause and also gets people engage in them, made as part of Database Management System course project.

  • Prison Management System: A software from my OOP course, which was created with the motivation to digitize the prison system of Bangladesh. (It was my first project)

See more of my projects on Github

Certifications