Skip to content
View ishan121028's full-sized avatar

Block or report ishan121028

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ishan121028/README.md

Ishan Upadhyay

Member of Technical Staff at Adobe | IIT Bombay, Class of 2025

Personal website | Resume (PDF) | LinkedIn | Email

I build cross-platform software and developer tools for Creative Cloud Desktop. Alongside product engineering, I collaborate on research in human-AI interaction, multimodal evaluation, and model calibration.

Engineering

At Adobe, my work spans C++ platform components, Windows/macOS integrations, and developer tooling.

  • Built Patrol, a production-debugging system with evidence-backed investigations and human-reviewed fixes: 100+ issues resolved at a 60-70% end-to-end resolution rate.
  • Implemented incremental feature-flag caching that saved approximately $200,000 in service costs.
  • Delivered an Adobe Express Photos cross-platform prototype in under five weeks.

Recipient of Adobe's Special Contribution Award and named inventor on a browser-to-desktop communication patent application, with filing in progress.

Selected projects

  • RadiologyAI: A software prototype combining LandingAI document extraction with Pathway streaming indexing and REST/MCP retrieval for radiology reports.
  • UniTrain: Contributor to an open-source deep-learning library developed through Hello FOSS at IIT Bombay.

Research

  • GRACE, ACL 2025. Co-author; built the question-authoring platform used by 100+ writers for a benchmark of human and model calibration.
  • AlphaCross-XL, Molecular & Cellular Proteomics, 2025. Co-author; developed the interface and computational workflows for protein cross-linking analysis.
  • Multimodal quizbowl (submitted to NAACL): Research collaboration with Prof. Jordan Boyd-Graber at the University of Maryland. Built authoring and review pipelines supporting iterative text-and-image drafting and model feedback. Manuscript: Asking About Occluded Content Marks Adversarial Multimodal Questions.

Writing

Building a Real-Time Radiology AI System with Pathway and LandingAI - Pathway

Background

B.Tech. in Aerospace Engineering, IIT Bombay (2025). Previously Institute Manager for the Web and Coding Club and a CS101 teaching assistant.

Tools I work with: Python, C++, TypeScript, React, Node.js, FastAPI, PyTorch, and CEF.

Pinned Loading

  1. wncc/UniTrain wncc/UniTrain Public

    UniTrain is an open-source, unified platform for effortless machine learning model training, evaluation, and deployment across diverse tasks. Experience seamless experimentation and model deploymen…

    Python 18 23

  2. RadiologyAI RadiologyAI Public

    Streaming radiology-report processing with LandingAI extraction, Pathway indexing, and REST/MCP retrieval. Software prototype.

    Python 4 2

  3. duckdb/duckdb duckdb/duckdb Public

    DuckDB is a high-performance analytical database system

    C++ 42.7k 4.1k