Md Atik Ahamed

Md Atik Ahamed

Postdoc@Stanford

About Me

I am a Postdoctoral Scholar at Stanford University. I completed my PhD in Computer Science at the University of Kentucky. My research lies at the forefront of deep learning, multimodal learning, and real-world data challenges to build highly effective AI solutions. During my PhD, I interned at Google (≈9 months), leading the development of multi-agent systems and foundational architectures like TFRBench (ICML 2026) and STRIDE. I was also selected for the exclusive Meta PhD Forum.

My research has achieved broad impact across both industry and academia, leading to publications in venues such as ICML, AAAI, KDD, ECAI, Nature Communications, Information Fusion, Journal of Biomedical Informatics, and Medical Image Analysis.

Previously, I earned my BSc from RUET and lectured at Green University of Bangladesh. As a Kaggle Competition Expert, I also enjoy solving practical ML problems. I’m always interested in new ideas and exciting collaborations!

Recent News

Paper Regarding Molecule Generation accepted to Nature Communications!
June 2026
Details coming soon.
Reasoning-Aware Training for Time Series Forecasting
May 2026
Preprint released @Google Internship
Novel framework integrating LLM reasoning natively into the continuous embedding space.
arXiv
TFRBench: A Reasoning Benchmark for Evaluating Forecasting Systems
April 2026
ICML 2026! @Google Internship
1st reasoning benchmark for time-series forecasting.
Paper Website
RefiDiff: Progressive Refinement Diffusion for Efficient Missing Data Imputation
Nov 2025
AAAI 2026! (Acceptance rate for year 2026 < 18%) Unifies predictive and generative paradigms for robust and flexible imputation.
Paper Code
Selected to participate in the Meta PhD Forum, 2025
Oct 2025
(sponsored by Meta)
Post
CausalGeD: Blending Causality and Diffusion for Spatial Gene Expression Generation
May 2025
KDD 2025! (Acceptance rate < 19%) Novel blending of diffusion and AR in Spatial Transcriptomics.
Paper Video Code
TSCMamba: Mamba meets multi-view learning for time series classification
March 2025
Accepted by Information Fusion! (Impact Factor 14+) Robust time-series classification using multi-view learning.
Paper Code
Low-dose computed tomography perceptual image quality assessment
January 2025
Accepted by Medical Image Analysis! (Impact Factor 10+) First low-dose CT IQA challenge.
Paper Code
Timemachine: A time series is worth 4 mambas for long-term forecasting
July 2024
Accepted to 27th ECAI (Acceptance rate < 24%) Unifies channel-mixing and channel-independence.
Paper Code
MambaTab: A Plug-and-Play Model for Learning Tabular Data
May 2024
Accepted to 7th IEEE MIPR (Acceptance rate < 20%) Extremely small model size.
Paper Code

Contact

Feel free to reach out using the following email (remove *):

Stanford, CA, USA
a*t*i*k*[at]stanford.edu