Vol. I — Portfolio Edition Est. 2026 Austin · New York

JANAVI KOLPEKWAR

Artificial Intelligence/Machine Learning · Data Science · Applied Research

USTA Innovation Hackathon 🏆 1st of 43 teams NKF Spring Clinical Meeting Best Poster of 289 Hack Texas 2025 BrainBoard — built in 24h Published Researcher CytoJournal · Stanford Bioengineering

About

Janavi Kolpekwar
ML Engineering Intern, Siemens EDA — Calibre

I currently build physics-informed ML simulation pipelines for semiconductor lithography, used at production scale by clients like Intel and Samsung.

Columbia University — MS Data Science, 2027

UT Austin — BS Statistics & Data Sciences, 2026

LinkedIn ↗ GitHub ↗ Email
Janavi Kolpekwar

What I Work On

I.

Physics-informed ML for chip manufacturing

Simulation pipelines with physics-informed kernels for semiconductor lithography, running at production scale for clients like Intel and Samsung.

II.

Agentic frameworks that do real work

Multi-agent and RAG systems that automate real workflows — from documentation to live data analysis.

III.

Turning speech into structure, live

Speech-to-text and NLP pipelines feeding real-time graphing, medical scribing, and mind-mapping systems.

Experience

Siemens EDA — Calibre

ML Engineering Intern

June 2025 — Present

Simulation. Physics-informed ML pipelines for chip lithography — cut critical-dimension margins by 30%.

Accuracy. Improved aerial-image correction by 38% for advanced chip manufacturing nodes.

Scale. 50× throughput gain in aerial-image generation — jobs down from 2 hours to under 3 minutes — for clients including Intel and Samsung.

Selected Work

Feature — Hack Texas 2025

BrainBoard: a conversation, mapped as it happens

Live conversation, transcribed and mapped into a real-time, searchable mind map — with Google Scholar lookup on every node. Built end-to-end in 24 hours.

Python · React · Deepgram STT · AWS DynamoDB / Lambda · Scholar API

View project ↗

AI/ML Research Analyst

Navigo Health Systems: NLP for a medical AI scribe

NLP algorithms and MVP work for a medical AI scribe, with speech-to-text pipelines cutting manual preprocessing effort 40% and enabling faster iteration cycles.

Python · AWS HealthScribe · Deepgram · Nabla

Research & Publications

01 MMRct — a non-invasive metric for coronary artery disease. Taylor C, Kolpekwar JA. View repo material ↗ Stanford Bioengineering · Apr 2025 – present
02 IHC Data Interpreter with GPT-4o for diagnosis of metastatic cancer sources. Kolpekwar JA, Farahi A, Shidham VB. View repo material ↗ Sept 2025
03 Effects of music therapy on dialysis patients: a quality improvement project. Kolpekwar JA, Birewar SP, Shanks M, Garcia S, Weber S. Read ↗ NKF Spring Clinical · Best Poster of 289 · Mar 2022
04 Impact of cytopathology author's work: open-access vs. non-open-access publications. Kolpekwar JA, Shidham VB. Read ↗ CytoJournal 2021;18:20 · Aug 2021

Toolkit

LanguagesPython · R · SQL · C++ · JavaScript (React/TS) · Go · Qualtrics XML
ML / DataPyTorch · TensorRT-LLM · Databricks · Apache Spark · Airflow · Streamlit
SystemsConcurrency · micro-batching · benchmarking · throughput optimization
Cloud / InfraAWS (DynamoDB, Lambda, REST) · GCP · PostgreSQL · Hadoop · Collibra · PowerBI
WorkflowGit/GitHub · Jira · Linux/Unix · CI