How I think
I am most comfortable in the middle ground between biology and engineering. I care about the mechanism, but I also care about whether the spreadsheet, split, figure, and code actually support the claim.
Student researcher + computational biology
Aarnav Bhat
I am a student researcher drawn to the places where biology becomes measurable: fragments of DNA in blood, methylation changes across cancer cohorts, cell states in tumors, and protein structures that can be tested instead of just imagined.
The thread running through my work is validation. I like building models, but I am more interested in whether they survive the uncomfortable parts: clinical controls, held-out data, ablation studies, negative controls, technical confounders, and clear limits on what the results can prove.
computational oncology
machine learning
protein modeling
public datasets
scientific writing
My goal is to become the kind of researcher who can connect biological intuition, reproducible code, careful validation, and honest communication without treating any one of those as optional.
About me
I like building things that force ideas to become testable.
I started from a curiosity about disease biology and kept getting pulled toward the computational side because it lets a question become a pipeline: find the data, clean it, decide what a fair test would be, run the model, and then ask what could have fooled it.
What I am training toward
I want to contribute to computational genomics, cancer diagnostics, protein modeling, and translational research where models are useful because they are tested against the messy parts of biology.
Outside the code
I also care about communication and discipline: public science writing, Sanskrit teaching, competitive swimming, and learning how to make hard ideas easier for other people to enter.
computational biology
cancer genomics
machine learning validation
protein modeling
scientific writing
reproducible research
AlphaFold
Boltz
ESM models
Python pipelines
scikit-learn validation
computational biology
cancer genomics
machine learning validation
protein modeling
scientific writing
reproducible research
AlphaFold
Boltz
ESM models
Python pipelines
scikit-learn validation
AlphaFold
Boltz
ESM models
Python pipelines
scikit-learn validationHow I work
The part I like most is turning a vague idea into a fair test.
My favorite problems sit at the point where biology, data, and communication all matter at once. I am trying to train myself to be curious without being careless, ambitious without overclaiming, and technical without losing the biological question.
Biology first
I try to start with a mechanism or clinical question before touching the model. A clean metric is less meaningful if the biological setup is weak.
Validation as a habit
I like held-out tests, repeated splits, negative controls, ablations, and clear limitations because they make the result harder to fool.
Structure and sequence
I am increasingly interested in how sequence models and protein-structure tools can connect molecular hypotheses to testable designs.
Readable science
I care about writing because a project is not really finished until another person can understand what was done and what was not proven.
Research timeline
A focused path into computational oncology.
Validation-first biomedical machine learning
Deepened my work in cancer genomics, patient-level validation, external checks, ablation studies, PR-AUC, confusion matrices, negative controls, and technical-confounder audits.
Epigenomics and cancer diagnostics
Studied how methylation, expression, and cohort-scale signals can be translated into supervised learning problems without losing sight of clinical context.
Wet-lab genetics foundation
Studied DNA damage and repair concepts through micropipetting, DNA extraction, and gel electrophoresis, with formal public recognition for the project.
Coursework and communication
Continuing work in precision medicine, genetics, machine learning, public science writing, Sanskrit teaching, and competitive swimming.
Writing and manuscripts
Research communication without overclaiming.
I am learning that clear writing is part of the science. It forces me to separate what the evidence supports from what I hope the next experiment might show.
Technical toolkit
Tools I use to turn messy biomedical data into testable models.
Python
pandas
NumPy
scikit-learn
Feature engineering
ROC/PR analysis
Permutation testing
SHAP
LaTeX/Overleaf
Git/GitHub
cfDNA
Epigenomics
Spatial validation
Structure-aware screening
Contact
Open to remote computational research support.
Interested in genomics data analysis, literature review, reproducibility checks, model validation, and manuscript support during the fall or winter.