Shaun Porwal

Metuchen, NJ shaun.porwal@gmail.com shaunporwal.com github.com/shaunporwal cheminventory.co

Experience

Founder & EngineerJunto Technologies
- Built a full-stack (design, client, server, database, authentication) chemical inventory management system (cheminventory.co) that tracks $1M+ in yearly chemical order volume with full regulatory compliance and version-controlled document provenance
 
- Architected a self-hosted, autonomous multi-agent AI swarm (Tailscale device fleet, local LLMs via vLLM/llama.cpp, shared task queue, persistent memory, self-improving skills, cron-scheduled runs, promptfoo evals, custom monitoring dashboard, Apple Watch voice control) to run market research, competitive analysis, and go-to-market strategy — delivering a full ICP/GTM report and verified competitor benchmarks
 
- Standardizing processes at a specialty chemicals company to achieve ISO certification
 
- Served as Mandarin/English translator at CPHI Shanghai 2026 for a specialty chemicals company
Founding EngineerStandard Model Biomedicine
- Architected a full-stack agent orchestration platform (FastAPI/Next.js, WebSockets) for biomedical workflows, integrating OpenAI Agents SDK tooling and DICOM/NIfTI imaging for tumor analysis
 
- Built a reusable research library (smb-pipe) for survival-prediction pipelines, automating CT embedding extraction and MEDS-to-model workflows for aortic aneurysm and heart failure risk modeling
 
- Developed an LLM benchmarking framework for clinical trial matching, implementing vLLM evaluation runners and custom embedding classifiers (MLP / logistic regression heads) on EHRSHOT data
 
- Secured a strategic data-for-equity partnership with Memorial Sloan Kettering Cancer Center, facilitating access to high-fidelity multimodal datasets for foundation model development
Machine Learning Engineer / Data AnalystMSKCC
- Engineered SQL-sourced data pipelines and statistical code for Amplio, enabling sarcoma, melanoma, Whipple, gastrectomy, and liver surgeons to benchmark patient outcomes against peers
 
- Built an AI radiology pipeline (Llama 3.1, SAM image segmentation, RShiny dashboard) to assist radiologists
 
- Led statistical analyses (GAM, logistic regression, Kaplan-Meier, CoxPH, Decision Curve Analysis) for a landmark study on post-chemotherapy RPLND policy in testicular cancer, published in Annals of Oncology; contributed statistics to additional publications on prostate cancer and robot-assisted surgery outcomes
 
- Developed dcurves, a Python package for Decision Curve Analysis (72k+ PyPI downloads); built and maintain the instructional site decisioncurveanalysis.org
 
- Engineered a Llama 3.1:70B + RAG text-to-SQL tool (no-more-sql) with Streamlit and FAISS for natural-language querying of MSKCC's SQL databases
 
- Completed a large patient-comorbidity analysis using take-home survey data, with pointblank data validation, to demonstrate accurate billing-level assignment
Bioinformatics InternSema4
- Engineered automated Bash/Python/WDL/AWS pipelines to detect structural variants across 100+ 100GB+ BAM files using Delly, SvABA, Manta, and Illumina DRAGEN
 
- Algorithmically discovered repeat variants among 100+ samples, comparing insulin- vs. non-insulin-producing pancreatic neuroendocrine tumors for diabetes therapy research
 
- Visualized results with Circos plots, stacked bar plots, heat maps, and scatter plots
Inorganic Manufacturing AssociateSPEX CertiPrep
- Performed inorganic standards analysis via wet-lab techniques (pipetting, gravimetry, titration, filtration, combustion) and ICP-MS instrumentation for reference-standard preparation
 
- Programmed and calibrated a Hudson automated pipettor robot (SOLOsoft) for serial dilutions and standard synthesis
Quality Control InternTokyo Chemical Industry Co., Ltd
- Trained 160 hours in instrumentation (GC, HPLC, NMR, IR) and wet-lab QC analysis
 
- Assisted with Japanese/English translation for the team

Selected Projects

dcurves
Python package for Decision Curve Analysis, evaluating the clinical value of prediction models and diagnostic tests; 72k+ PyPI downloads and used in peer-reviewed research.
no-more-sql
Llama 3.1:70B + RAG pipeline (Streamlit, FAISS) for natural-language querying of SQL databases without writing SQL.
cheminventory.co
Full-stack chemical inventory management and regulatory compliance platform built for Junto Technologies.

Education

M.S., Biomedical Data Science
Icahn School of Medicine at Mount Sinai
B.S., Biomedical Engineering & Chinese
Rutgers University–New Brunswick
Huayu Enrichment Scholarship, Taiwan MOE
Certificate in Advanced Mandarin Chinese
National Taiwan University
MOE Scholarship recipient

Skills & Certifications

Programming
Python, R, TypeScript, SQL, Bash, HTML/CSS, Docker, Git
AI / ML
LLM agent orchestration, multi-agent/swarm systems, local LLM deployment (vLLM, llama.cpp), self-hosted GPU/edge infra (Tailscale), LLM evals (promptfoo, Argilla), autonomous scheduling, RAG, fine-tuning, image segmentation, embedding classifiers
Statistics
GAM, CoxPH, Kaplan-Meier, logistic regression, Decision Curve Analysis, survival analysis
Domain
Chemical regulatory compliance / ISO, bioinformatics (SV detection, RNA-seq)
Languages
English (native); Mandarin Chinese, Hindi (fluent); Japanese, Spanish (conversational)
Certifications
Epic – Clinical, Epic – Reporting, Advanced Mandarin Chinese (NTU)