FounderRecord

Raunak Chowdhuri

Co-Founder · CTO · San Francisco

Indian-American co-founder and CTO of Reducto, an AI document intelligence platform that has raised $108M and processes over one billion pages per month. Previously co-founded Oloren AI; published ML research with 100+ citations before finishing high school.

Based in San Francisco, CaliforniaLast reviewed October 2026
Raunak Chowdhuri
Photo: GitHub (raunakdoesdev)
Compiled from public sources. Raunak Chowdhuri has not yet contributed to this record.
Claim your record

Overview

Raunak Chowdhuri is an Indian-American computer scientist and entrepreneur serving as co-founder and chief technology officer of Reducto, a San Francisco–based AI document intelligence platform. Reducto converts unstructured documents into structured, machine-readable data for use in large language model pipelines, and has raised $108 million in venture funding at a valuation of approximately $600 million.[1][18][4]

Chowdhuri began publishing machine learning research at age 15, accumulating over 100 academic citations before enrolling at the Massachusetts Institute of Technology in 2020.[4][20] During his undergraduate years he held concurrent research positions at MIT CSAIL, MIT Lincoln Laboratory, Carnegie Mellon University, Caltech, and Nvidia, and co-founded Oloren AI, a computational chemistry startup that shut down in 2023.[22] He co-founded Reducto later that year with MIT classmate Adit Abraham after a hackathon project revealed that document ingestion was a critical unsolved bottleneck in AI workflows.[4][7] In 2026, both founders were named to the Forbes 30 Under 30 list in the Artificial Intelligence category.[1]

Early life and education

Chowdhuri (legal name Sauhaarda Chowdhuri) grew up in San Diego, California, and attended Westview High School, where he graduated in 2020.[17] He began computer vision research at age 12 and published his first academic paper at approximately age 15.[4][17] His early research affiliations included positions at UC Berkeley (2017), Rockley Photonics (2016), and IEEE-RAS (2018).[17] In 2019, he won multiple awards at the Intel International Science and Engineering Fair, including the GoDaddy Best Use of Data Award and third place in Robotics and Intelligent Machines, and received a $10,000 Kyoto Prize Symposium Scholarship from the Inamori Foundation.[17]

Chowdhuri enrolled at MIT in 2020 and graduated in May 2024 with a Bachelor of Engineering in Computer Science and Electrical Engineering.[25][11] In his freshman semester, he served as a Learning Assistant for a graduate-level machine learning course composed mostly of PhD students — a role made possible by his pre-college publication record of over 100 citations.[4][7] It was in that course that he first met Adit Abraham, then a junior, who would become his co-founder at Reducto.[4]

Career

Early Research — Rockley Photonics, UC Berkeley, and IEEE-RAS (2016–2020)

While still in high school, Chowdhuri conducted machine learning and computer vision research across several institutions. He held research positions at Rockley Photonics (2016), UC Berkeley (2017), and IEEE-RAS (2018).[17] This pre-college research phase produced two notable publications: "MultiNet: Multi-Modal Multi-Task Learning for Autonomous Driving," co-authored with Tushar Pankaj and Karl Zipser and submitted to arXiv in September 2017 when Chowdhuri was approximately 15 years old, later published at IEEE WACV 2019;[12] and PhonoNet, a hierarchical deep learning system for raga identification in Hindustani classical music that achieved 99% accuracy across more than 30 ragas, published at ACM ICMR 2019.[13] By the time he finished high school, his papers had accumulated over 100 academic citations.[17][20]

MultiNet — Autonomous Driving Research (2017)

Chowdhuri's most-cited early work, "MultiNet," introduced a multi-modal, multi-task learning approach for end-to-end autonomous driving, trained on hours of driving data from 1/10th-scale model cars.[12] The paper, published under his legal name Sauhaarda Chowdhuri, has accumulated 137 academic citations as of 2026.[20] The work was conducted in collaboration with UC Berkeley researchers and represented one of the earliest demonstrations of multi-task neural networks applied to autonomous vehicle perception.[12]

MIT Research Positions — CSAIL, Lincoln Laboratory, MIT Driverless, Carnegie Mellon, Caltech, and Nvidia (2020–2023)

During his undergraduate years at MIT, Chowdhuri held an unusual number of concurrent research and industry positions. He served as a Research Assistant at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) from 2020 to 2021, worked as a Machine Learning Research Consultant at MIT Lincoln Laboratory, and contributed to MIT Driverless as a perception team researcher applying deep learning to autonomous vehicle sensing.[17] He also held research roles at Carnegie Mellon University and Caltech, and interned at Nvidia as an Autonomous Driving Research Intern in summer 2021.[17] Additional roles during this period included work at DeGirum Corp (2020), Peptide Logic (2020–2021), and the MIT Entrepreneurship Review (2020–2021).[17]

GPT-4 Rebuttal — MIT Independent Research (2023)

In June 2023, Chowdhuri gained notoriety at MIT for debunking a viral research paper that claimed GPT-4 could score 100% on MIT's EECS curriculum. Working alongside Neil Deshmukh and David Koplow, he published a detailed rebuttal proving the authors' evaluation methods were invalid.[17] The paper was subsequently retracted, and the episode eventually led to a professor's dismissal, according to Contrary Research.[17] The investigation established Chowdhuri's reputation for rigorous technical scrutiny within the AI research community before he had completed his undergraduate degree.[17]

Oloren AI (2021–2023)

Chowdhuri co-founded Oloren AI in 2021, a San Diego–based computational chemistry startup building an AI-based molecular property prediction platform for drug discovery.[22] As CTO, he led development of Oloren ChemEngine (OCE), an open-source Python library providing a unified API for molecular property predictors, enabling AI-driven drug discovery teams to share, distribute, and ensemble models with simplified reproducibility.[23] He co-authored the foundational ChemRxiv paper on the platform in October 2022 and presented at the Drug Discovery Chemistry conference in San Diego in April 2022.[23][23] Oloren AI raised approximately $300,000 in angel funding from AB Magnitude Ventures Group.[22] The company went out of business in September 2023, at which point Chowdhuri pivoted to Reducto, which he had already begun building.[22][17]

Separately, Chowdhuri also founded and scaled an independent computational chemistry consulting practice to $200,000 in annual recurring revenue during his time at MIT.[17]

Reducto (2023–present)

Chowdhuri co-founded Reducto in 2023 with Adit Abraham, whom he had met in a graduate-level ML course at MIT where Chowdhuri was the freshman Learning Assistant.[4][7] The two joined the same MIT living group and were formally paired in a mentorship program, deepening a friendship that predated any business plans.[7] After Oloren wound down and Abraham grew restless at Google, they began entering hackathons together. They won Anthropic's hackathon at the first release of Claude, building a product that let users upload PDFs and manage chat context.[7] When they interviewed potential customers, the document upload feature drew far stronger interest than the chat functionality — revealing that document ingestion, not retrieval, was the real bottleneck in AI workflows.[4]

The pair applied to Y Combinator with a long-term memory product for language models and were accepted into the Winter 2024 batch. One day after acceptance, they told their YC partner Diana Hu that they did not believe the memory product was the right direction, and pivoted to document parsing.[7] Chowdhuri's early technical breakthrough was reinventing a 1970s algorithm called XY Cut — a simple method for ordering and filtering document regions — which outperformed enterprise OCR systems from Google and Amazon because it was built with total focus on the parsing problem.[4] The company's name is a reference to the Reducto Charm from Harry Potter, a spell that blasts solid objects into pieces — analogous to breaking documents into clean, structured data.[18]

The thing that finally worked was reinventing an algorithm from the 1970s called XY Cut. It was simple and naive, just ordering and filtering regions. But it performed significantly better than the more advanced systems from Google or Amazon because it was built with total focus on the problem.Raunak Chowdhuri · 2026

While still a two-person YC team, a Fortune 10 company contacted Reducto after its parsing tool successfully decoded one of their complex documents out of the box.[4] The company reached seven-figure annual recurring revenue within six months of launching, crossing $1 million ARR with only four employees.[23] Reducto reached $5 million in revenue before hiring its first salesperson.[17]

Reducto raised an $8.4 million seed round led by First Round Capital in October 2024, with participation from Y Combinator, SV Angel, Liquid 2 Ventures, and angel investors including Dropbox co-founder Arash Ferdowsi.[17][4] In April 2025, the company raised a $24.5 million Series A led by Benchmark.[18] Five months later, in October 2025, Reducto closed a $75 million Series B led by Andreessen Horowitz, bringing total funding to $108 million at a post-money valuation of approximately $605 million.[19][18][4] The company had burned only $1 million in capital from inception to the Series B raise.[17]

By the Series B announcement, Reducto had surpassed one billion pages processed cumulatively, with monthly processing volume growing more than 6x in the five months between the Series A and Series B.[20][25] Named customers include Scale AI, Harvey, Rogo, Airtable, Vanta, and an undisclosed Fortune 10 enterprise.[18][22] In September 2026, Reducto launched r-1, a proprietary document parsing model combining OCR with vision-language models to handle complex layouts, tables, handwriting, and citations.[23] In May 2026, the company made its first acquisition, purchasing Y Combinator–backed edtech startup Opennote and absorbing its entire team.[24] As of mid-2026, Reducto had approximately 42 employees.[17]

Public statements

Documents contain some of the most valuable data in most industries — from healthcare to finance to logistics. Yet until now, they've been a bottleneck for making AI useful for real enterprise use cases. Our vision is to build the trusted layer that connects messy, real-world data with language models — so that AI can reason over the world as it really is, not just the text that's clean and easy.

On Reducto's mission — Pulse 2.0, October 2025[20]

When we started building Reducto, I tried every 'enterprise grade' OCR and parsing library out there. None of them could reliably handle the tougher, production grade documents we were seeing.

On the technical gap that motivated Reducto — LinkedIn, 2026

Personal life

Chowdhuri is of Indian-American heritage and grew up in San Diego, California.[25][17] He goes by "Raunak" professionally; his academic publications appear under his legal name, Sauhaarda Chowdhuri.[12][17] He has a personal interest in Indian classical music, reflected in his PhonoNet research project on Hindustani raga identification.[13] He resides in San Francisco, California.[17]

Media appearances

References

  1. Reducto — Forbes 30 Under 30 2026 Profile — Forbes, December 2025
  2. Investing in Reducto — Andreessen Horowitz (a16z), October 2025
  3. Reducto Valuation, Funding & Investors — Multiples.vc
  4. From Weekend Project to Fortune 10 Adoption — Reducto's Path to Product-Market Fit — First Round Capital Review, April 2025
  5. Raunak Chowdhuri — Google Scholar — Google Scholar
  6. Oloren AI - 2026 Company Profile, Funding & Competitors — Tracxn
  7. How a weekend hack became a multimillion-dollar AI startup | Adit Abraham (Co-founder & CEO at Reducto) — First Round Capital Review (Podcast), April 2025
  8. Reducto Business Breakdown & Founding Story — Contrary Research, August 2026
  9. Reducto: The complete agentic document platform — Y Combinator
  10. MIT Commencement 2024 Graduates Name Book — MIT, May 2024
  11. Look Who from MIT Made the Forbes Under 30 List — MIT Alumni Association, December 2025
  12. MultiNet: Multi-Modal Multi-Task Learning for Autonomous Driving — arXiv / IEEE WACV 2019, September 2017
  13. Sauhaarda Chowdhuri — DBLP — DBLP
  14. A Unified System for Molecular Property Predictions: Oloren ChemEngine and its Applications — ChemRxiv, October 2022
  15. Artificial Intelligence for Early Drug Discovery — Drug Discovery Chemistry, April 2022
  16. Reducto AI: Document Parsing for LLM Pipelines — IDP-Software
  17. Reducto Raises $24.5M in Series A Funding — FinSMEs, April 2025
  18. Exclusive: Reducto, AI document parsing startup, raises $24.5 million Series A led by Benchmark — Fortune, April 2025
  19. Reducto Raises $75M in Series B Funding — FinSMEs, October 2025
  20. Reducto: $75 Million Series B Secured For Advancing AI Document Intelligence Expansion — Pulse 2.0, October 2025
  21. Reducto Raises $75M Series B to Define the Future of AI Document Intelligence — PR Newswire, October 2025
  22. AI Document Intelligence Platform Reducto Secures $75M Series B — Built In SF, October 2025
  23. Reducto launches r-1 at 1 cent per page for document parsing — Runtime Wire, September 2026
  24. Canadian-founded Opennote acquired by San Francisco-based Reducto — BetaKit, May 2026
  25. 50 Indian Americans among Forbes' 2026 30 Under 30 changemakers — American Bazaar Online, December 2025

Is this your record?

Claim it to review for accuracy, or enrich it through a personal interview.

Claim your record
founderrecord.com/raunak-chowdhuri
FR–2026–0090·Updated October 15, 2026