// premise
Is the drug target real?
I build open-source computational pipelines that interrogate whether a drug target is actually real — and try as hard to falsify a hypothesis as to confirm it. Every project ships a pre-registered hypothesis, content-addressed (SHA-locked) verdicts so results can't quietly drift, and null results published next to the positive ones. Most of it runs end-to-end on free Kaggle GPUs.
pre-registered hypotheses · content-addressed verdicts · gaps documented, not smoothed · free compute
// now
What I'm working on
Current, active lines of work.
Nav1.7 (SCN9A) — the genetic–pharmacological asymmetry of a "perfect" pain target
A full analysis of why lifelong genetic loss of Nav1.7 abolishes pain at no cardiovascular cost, yet acute pharmacological block keeps failing in the clinic — biophysics, expression, and human-genetics arms converging on a two-sided constraint. Shipping as a reproducible pipeline and write-up: crisprking/nav17-asymmetry.
Medical school — MS1, UAG International MD
Training clinically while keeping an active open-source computational research line.
Open drug-target auditing
Extending falsifiable-targets — pre-registered, SHA-locked verdicts on whether a target's genetics actually support the proposed direction of effect.
// selected work
Frameworks & methods
Reusable tools for deciding whether a target — or a model's claim about it — can be trusted.
A content-addressed audit engine for drug-target direction-of-effect (inhibit vs. activate), built on Open Targets colocalization and benchmarked against approved drug–target pairs — with the failure modes it can't yet handle documented rather than hidden. The production layer adds batch auditing, Nextflow + Snakemake DAGs, and a Docker image.
known gaps: 3 documented failure modesWhen can you trust ESM-C's zero-shot variant-effect rankings — and when can't you? A small, reproducible benchmark showing that a model's self-consistency across sizes does not predict its accuracy, plus a calibration tool for deciding when to rely on it.
Why the best genetically validated pain target keeps failing in the clinic. Makes the case for a genetic–pharmacological asymmetry: lifelong genetic loss of Nav1.7 is cardiovascular-silent, while acute pharmacological block causes on-target autonomic toxicity. Single-cell atlas analysis, a leaky homeostatic-compensation model, and gnomAD / ClinVar / Open Targets genetics, converging on three independent 2024 publications.
Target-discovery pipelines
Public data to a defensible shortlist, end-to-end, on free compute.
Maps type-1-diabetes GWAS loci to the pancreatic cell types they likely act in, using τ-based cell-type specificity in the HPAP scRNA-seq atlas, and cross-validates against autoantibody-positive pre-clinical transcriptional change.
Runs a standard ChEMBL-driven target pipeline on Madurella mycetomatis, a neglected fungal pathogen nobody had mapped — and catches the pipeline's own artifact: the "top" gene was really 384 duplicate records of HDAC4. Triaged to a hardened shortlist, with the audit that caught the error shipped alongside.
An eight-stage open drug-discovery pipeline for Chagas disease targeting the T. cruzi protease cruzain: three parallel scoring tracks fused into a consensus, a selectivity counter-screen against three human cathepsins, and ADMET filtering. Runs on a free Kaggle T4.
A structure-based paralog selectivity counter-screen for the immuno-oncology target ENPP1: dock candidate inhibitors into ENPP1 and its two close cousins ENPP2 (autotaxin) and ENPP3 with one identical zinc-centered box, then rank by cross-paralog margin rather than raw affinity. The top affinity binder reversed to an off-target liability; native controls, bootstrap CIs, and two honest negative benchmarks ship alongside. Runs on a free Kaggle T4.
An empirical evaluation of which public-data features actually predict PROTAC E3-ligase tractability.
Tools
ncbi-bioscraper (zero-cost PubMed + OpenAlex + open-access full-text mining), target-confidence-card, miniprotein_genai, and an MCAT concept-practice app for premeds.
// background
Education
Biomathematics, bioinformatics, and computational biology.
Genetics, genomics & cell biology; immunochemistry and cell culture. Thesis: Refining the epigenome through CRISPR-mediated techniques to establish a programmable system for transcriptional memory.
Research & lab
-
2018–21
Pines Lab Undergraduate Researcher — UC Berkeley
Magnetic-resonance research in the Pines Lab: zero- to ultra-low-field (ZULF) NMR and imaging methods that remove the need for strong magnetic fields. Worked with laser-polarized xenon molecular sensors, solid-state NMR of NV-diamond materials, optical hyperpolarization for signal enhancement, and miniaturized NMR detectors for portable bioimaging.
-
2021–22
Medical Technician — Discover Labs
Real-time PCR panels and automated RNA extraction under CLIA/HIPAA, with antimicrobial-treatment guidance and QC-workflow improvements.
-
2022
Microbiology Technician — Varian
Assessed non-invasive cancer therapies and ran viable / non-viable cleanroom monitoring under GLP, with supporting data analysis.
-
2018
Research Assistant — Tecnológico de Monterrey (Centro del Agua)
Microalgae bioprocessing at the Centro del Agua water-research center: spirulina cultivation, phycocyanin and DHA microencapsulation, and a microalgae-based UV-protective cream.
Industry & commercial
-
2024–25
Inside Sales Specialist — EditCo Bio (CRISPR)
Sole inside rep for a CRISPR reagents company across 30 states — a bench-trained scientist supporting researchers through their experiments, solving protocol problems rather than only closing deals.
-
2022–23
Sales Account Manager — GenScript
Managed a synthetic-biology product portfolio and coordinated custom gene-synthesis projects across pharma, biotech, and academic accounts.
-
2022
Consultant — PSC Biotech (Moderna)
IQ/OQ equipment qualification, sterilizer SOPs, and cleanroom quality control for Moderna manufacturing equipment under FDA and SAP guidelines.
-
2020–22
Founder & CEO — Creative Science
Founded and ran a cross-border resale business for lab and medical equipment (US / Mexico), sourced from pharma and biotech auctions.
// how i work
Method
- ▸Pre-registered, falsifiable hypotheses — fixed before the analysis runs.
- ▸Content-addressed, SHA-locked verdicts, so a result can't silently change.
- ▸Refusal-first — the pipeline is allowed to say "not enough evidence," and gaps are documented rather than smoothed over.
- ▸Null results shipped next to the positive ones.
- ▸Reproducible on free compute (Kaggle T4), in self-contained notebooks.