Harsh Agrawal headshot

Harsh Agrawal

I’m interested in applying ML to solve interesting problems in biology and medicine. I’m currently part of SoftNanoLab, where I work on protein language models (PLMs). I am also pursuing an MEng in Molecular Bioengineering at Imperial College London.


Experience

Research Scientist · Softnanolab · In-Person
(London, UK)
  • Designing a Protein Language Model (PLM) specialized for protein complexes with Dr. Stefano Angioletti-Uberti at the Nanolab.
Research Scientist · Designing.bio · In-Person (London, UK)
  • Worked with Dr. Karen Sarkisyan & Andre Brown to design binders for C. elegans transmembrane receptors by repurposing protein toxins.
  • Implemented an end-to-end design pipeline by inpainting backbones with RFdiffusion, performing Rosetta relaxation, and running AlphaFold2 screening.
  • Designed a distributed inference stack across 40+ GPUs, scaling in silico throughput from ≈10 to >500 designs/day and yielding 10 candidates for experimental validation (out of 10K generated designs).
Lead Data Scientist · Velora · Remote
  • Built the public analytics pipeline and dashboard indexing 33M+ transactions and $140B+ in volume across 5 chains.
  • Migrated the data platform from Heroku to AWS, deploying Redshift + MongoDB for low-latency querying, resilient ingestion, and long-term storage.
  • Developed internal pipelines to simulate blockchain contract calls for route optimizations and the public DEX subgraph to index past trades.

Projects


Talks & Achievements