Mansoor Ahmed
Computer Science Graduate Student · Georgia State University / Georgia Tech
I am a graduate student in Computer Science at Georgia State University (joint with Georgia Tech), where I am advised by Prof. Murray Patterson at the Algorithmic Biology (AliBi) lab. My research focuses on generative ML for antibody design. I develop ML methods, such as graph neural networks, flow matching, diffusion models, and protein language models that condition on target antigens to design therapeutic antibodies. I also recently completed a research fellowship at UCLA's LUNE Lab under Prof. Jennifer Wilson.
Previously, I worked as a research assistant for one year at the Data Analytics lab at LUMS, Pakistan where I developed graph and Physics-based methods for antigen-antibody interaction, preprocessing pipelines for multi-modal MRI, and trained custom CV models for neurological disorders detection. Prior to that, I worked as a research assistant for one year at the Smart Data Systems & Applications (SDSA) lab at LUMS, where I developed the software and computer vision models for Vending Central that is now a leading vending company in Pakistan.
Research Interests: Generative models for protein design, equivariant & geometric deep learning, protein language models, antibody engineering, computational structural biology
Publications/Preprints
Mixture-of-Latents for Few-Step Discrete Diffusion Distillation
Overview
Latent-Kernel Discrete Diffusion for Few-Step Generation
Overview
AgForce Enables Antigen-Conditioned Generative Antibody Design (via annealed Multiple Choice Learning with Mixture Density Networks)
Overview
EpiFormer: Learning Antigen-Antibody Interactions for Epitope Prediction via Geometric Deep Learning
Interaction Maps
ChiMERa-Bench: A Benchmark Dataset for Epitope-Specific Antibody Design
Data Pipeline
Benchmarking Baselines
ConTact: Contact-First Antibody CDR Design via Explicit Interface Reasoning
Architecture
Key Results
EvoStruct: Bridging Evolutionary and Structural Priors for Antibody CDR Design via Protein Language Model Adaptation
Architecture
Amino Acid Recovery
Vocabulary Comparison
Improved Graph-Based Antibody-Aware Epitope Prediction with Protein Language Model Embeddings
Architecture
Results
Robust Brain Age Estimation via Regression Models and MRI-derived Features
Brain Aging
Pipeline
VBM Analysis
Teaching
CSC 8850 — Advanced Machine Learning, GSU [Course Page]
CS 3210 — Computer Organization & Programming, GSU
CSC 8713 — Spatial and Scientific Databases, GSU
EE 212 — Mathematical Foundations for Machine Learning, LUMS [Course Page]
Talks
1. AgForce Enables Antigen-conditioned Generative Antibody Design
2. Bridging the Gap Between LLMs and LNS with Dynamic Data Format and Architecture Codesign
3. An Hierarchical and Relation-aware Graph Neural Network for Antigen-Antibody Interaction Prediction
Honors & Service
Fellowships & Awards
- Molecular Basis of Disease (MBD) Fellowship, GSU 2025 – present
- Warren Alpert Computational Biology & AI Network Fellow, UCLA 2025 – 2026
- MBD and Martin Fraser travel awards to attend ICLR in Rio 2026
- Computational Genomics Summer Institute (CGSI), UCLA 2025
- Summer School in Biomedical Engineering, Max Planck / TU Ilmenau 2023
- HEC Merit Scholarship (fully funded BSc EE), Pakistan 2017 – 2021
- Balochistan Educational Endowment Fund 2015 – 2017
Reviewing & Service
- Workshops: GenBio @ ICML 2026, GEM @ NeurIPS 2026, ML4H @ NeurIPS 2026, BeNTO @ NeurIPS 2026
- Conferences: ISBRA 2025; 2026, ICCABS 2025, AAAI 2027
- Journals: PeerJ Computer Science, TMLR