scalable_learning.sys

phd researcher & research engineer
[dkfz & embl heidelberg]

Scalable Graph Learning
Spatial Inductive Biases
Multi-Omic Foundation Models
Distributed ML Infrastructure
Elyas Heidari
current_location

heidelberg, germany (49.4° n, 8.7° e)

specialization

ai for science / geometric dl

research_philosophy

i build learning systems that transform messy, large-scale biological data into reliable infrastructure.

my focus is designing gnn and transformer pipelines that scale to 10m+ nodes while preserving mechanistic interpretability.

[view_detailed_bio]
latest_updates
  • segger_framework v1.0 released
  • spatial_data_standardization update

scientific_dimensions

spatial & single-cell omics

end-to-end pipelines for spatial transcriptomics and multimodal integration.

scalable graph neural networks

multi-gpu training for heterogeneous gnns on graphs with 100m+ edges.

open-source tooling

architecting frameworks (segger, sagenet, scgcn) for biological discovery.