Aug 2022 – May 2026
Berkeley, CA
Garv Goswami
Student at UC Berkeley, learning my way around machine learning and the life sciences.
I’m an undergrad in CS, mostly just trying to be useful and keep learning. Along the way I’ve been lucky to work as a machine learning intern at Phare Bio and as a student researcher in the Ronda Lab at the Innovative Genomics Institute, where I help out on generative models for protein and small-molecule design.
i.
Education
Awards
- Warren Y. Dere Design Award · UC Berkeley EECS · 2026
Awarded by the Department of EECS to a graduating senior whose accomplishments in engineering design are judged most outstanding.
ii.
Industry Experience
Sep 2025 – Present
Boston, MA
Machine Learning Intern · Phare Bio
- Designing and deploying a scalable benchmarking platform on an HPC environment to evaluate molecular generative models — including JT-VAE, Transformer-based, and diffusion models — across large-scale runs (100k+ generated molecules).
- Conducted computational analoging efforts for small-molecule antibiotic programs using REINVENT4 and Chemprop, prioritizing candidate analogs based on predicted activity and developability-relevant properties.
Python / PyTorch / Small Molecule Design / Platform Development
May 2025
Mountain View, CA
Software Engineering Intern · Google
- Contributed to a 20% project exploring conversational recall for dementia patients using the Gemini Voice API; implemented dialogue flows and evaluated engagement to assess feasibility of AI-driven patient support.
- Collaborated with the YT Shorts Ads team to design and analyze new content and component placement for Shorts App Ads. Front-end in EML, server-side in C++. Used Google's internal experimentation platform to run gradual rollouts and A/B tests.
C++ / Experimentation / System Design
May 2024
Sunnyvale, CA
Software Development Engineering Intern · Amazon
- Created Integration and Load Testing CI/CD frameworks for new FireTV Ad Stack plugins, using Java to write tests on temporary Lambda compute to validate packages being added to the version set.
- Built a validation pipeline for FireTV Ads plugins entering Setu, Amazon's internal ad-customization tool. Programmed a system-monitoring tool in Java to track CPU, memory, and latency during plugin integration testing.
Java / System Design / AWS / Testing
iii.
Research Experience
Oct 2024 – Present
Berkeley, CA
Student Researcher · Innovative Genomics Institute — Ronda Lab
- Fine-tuned ESM-2 on CRISPR-family proteins to generate novel, compact variants of SpyCas-9 using a custom autoencoder model. Integrated AlphaFold for structural validation.
Protein Prediction Models / AlphaFold / Wet Lab Testing
Aug 2024 – Jan 2025
Remote Collaboration
First-Author — Clinical Trial Enrollment Prediction · UC Berkeley / Icahn School of Medicine at Mount Sinai
- First-authored draft with a Mount Sinai medical student on EnrollMate, a machine learning framework predicting clinical trial enrollment success.
iv.
Research Projects
Accepted to ICLR 2026 MLGenX Workshop
Contrastive Alignment of Expression and Copy Number Reveals Dosage-Insensitive Genes in Cancer
Built a contrastive learning framework aligning scRNA-seq expression and inferred CNV profiles in a shared latent space to identify dosage-insensitive genes in cancer, achieving 87%+ bidirectional top-1 retrieval accuracy on held-out cells.
PyTorch / Scanpy / infercnvpy
Read preprint (bioRxiv) →
Under Review at ISMB 2026
Fine-Tuning Genomic Language Models for Variant Pathogenicity Prediction
Benchmarked Nucleotide Transformer and Caduceus on ClinVar missense variant classification, showing variant-position embeddings outperform mean pooling and that LoRA reaches 0.886 validation AUC with only 0.76% trainable parameters.
PyTorch / LoRA / Transformers / Mamba
Preprint, 2025
EnrollMate — Clinical Trial Enrollment Prediction
First-authored preprint with collaborators at the Icahn School of Medicine at Mount Sinai introducing a machine learning framework for predicting clinical trial enrollment success.
Machine Learning / Healthcare
v.
Hackathon Projects
Neural Net Neutrality
An open-source project tracking how large language models' political leanings shift across versions, scoring responses to policy statements on economic and social axes and publishing reproducible time-series data.
Python / OpenAI API / Data Viz
View site →
Scrapbook
Collaborative scrapbooking app built at a hackathon — share moments with friends through a shared timeline.
React / Next.js / Node.js
Devpost →
LEED Bud
An OpenAI-powered assistant for navigating LEED green-building certification requirements.
React / TypeScript / OpenAI
Devpost →
TrashToTreasure
A sustainability marketplace pairing reusable materials with people who can give them a second life.
Next.js / MongoDB / Express
Devpost →
vi.
Get in Touch
I’m always happy to hear from collaborators, researchers, and students working on problems at the boundary of biology and machine learning.