Data Science · Machine Learning · Vibe Coding

Finding the non-trivial
in complexity.

Hi, I’m Abhishek. I turn raw and messy datasets into robust AI models that becomes production-grade framework.

Abhishek Mall
01 ABOUT
I

make difficult data
useful.

Developing methods for low-quality, high-dimensional datasets that break off-the-shelf tools.

  • AI : 3D Modeling, Representation Learning, Feature Extraction, and Continuous Optimization.
  • Systems : High-throughput Processing, Big-data Handing and End-to-End Pipelines.

Give me an ambiguous problem, petabytes of noisy data, and a high-stakes question — that’s where I build my best systems.

Python PyTorch SQL GenAI HPC / GPU
10⁶+ images analyzed
6 projects lead
200+ citations
02 EDUCATION

Built on first principles.

Physics taught me how to reduce complex systems to what matters. Computation taught me how to scale that thinking.

2020 — 2025 Hamburg, Germany
DOCTORATE

PhD
Physics

Universität Hamburg
International Max Planck Research School for Ultrafast Imaging & Structural Dynamics

THESIS

Unraveling Heterogeneity in X-ray Single Particle Imaging

2016 — 2020 Mumbai, India
DUAL DEGREE

MS
Physics

MTech
Materials Science

Indian Institute of Technology - Bombay

THESIS

Design and Optimization of Nanophotonic Metasurface using Deep Learning

2013 — 2016 New Delhi, India
UNDERGRADUATE

BS
Physics

Hansraj College, University of Delhi

FOCUS

Physics, Mathematical Methods and Computational Foundations

03 EXPERIENCE

From complex questions
to simple solutions.

I own the full arc: framing the question, designing the method, implementing the pipeline, scaling computation and communicating the result.

2020 — PRESENT
PH.D. RESEARCHER · HAMBURG

Max Planck Institute for the Structure and Dynamics of Matter

Built ML, Bayesian and optimization methods for large-scale X-ray imaging. Developed end-to-end analysis pipelines for millions of images and scaled computation across GPU and HPC systems.

Machine learning Bayesian analysis HPC
REMOTE
COLLABORATOR · SYRACUSE UNIVERSITY

Quantum Technology Laboratory

Developed a variational autoencoder with semi-supervised and transfer learning for robust classification under low photon statistics and noisy experimental conditions.

VAE Transfer learning Classification
2019 — 2020
RESEARCH ASSISTANT · MUMBAI

Indian Institute of Technology Bombay

Engineered deep-learning and genetic-optimization frameworks for fast inverse design of manufacturable nanophotonic devices.

Deep learning Optimization
2017
RESEARCH INTERN · PARIS

Laboratoire de Physique des Solides

Created a computational scattering model to study nanoscale self-assembly and interaction dynamics.

Modeling Numerical methods
04 SELECTED WORK

Proof, not
promises.

Research-grade problems approached with production-minded thinking.

01
UNSUPERVISED ML · LARGE-SCALE DATA

Finding structure in millions of diffraction images

Bayesian and machine-learning pipelines that classify, embed and characterize EuXFEL datasets—revealing reaction trajectories and structural changes in biological samples.

latent structure →
02
OPTIMIZATION · IMAGING

Optimization without reliable gradients

A noise-robust derivative-free algorithm for nonlinear holographic imaging problems.

03
DEEP LEARNING · DESIGN

Learning to design nanophotonic devices

A cyclical framework combining neural networks and genetic algorithms for inverse design.

05 RESEARCH

Peer-reviewed
evidence.

All publications → Google Scholar ↗
2024 Deep Learning based Variational Auto-Encoder for Classification of Quantum and Classical States of Light ↗ 2023 Holographic Single-Particle Imaging for Weakly Scattering, Heterogeneous Nanoscale Objects ↗ 2020 A Cyclical Deep Learning Framework for Inverse and Forward Design of Nanophotonic Metasurfaces ↗ 2020 Fast Design of Plasmonic Metasurfaces Enabled by Deep Learning ↗

Have a difficult data problem?

Let’s make it
understandable.

Connect ↗
GitHub LinkedIn Google Scholar