Parth Patwa

Timeline

1. 2010sCompletes a master's degree at the University of California, Los Angeles.Stated on his Y Combinator founder profile; the subject is not given there.^[4]
2. 2010s to 2020sPublishes machine learning research on fact checking, multimodal misinformation, meme sentiment and low resource language models, with papers at EMNLP, LREC and AAAI workshops.His indexed publication record.^[5]
3. 2020sWorks as a researcher at PathCheck Foundation, an MIT-linked public health software group listed among his previous employers.He describes this period as being a researcher at MIT.^[2]^[4]
4. 2020sJoins Amazon Web Services as a generative AI scientist.AWS published his author profile on a machine learning blog post about policy creation with Amazon Bedrock.^[4]^[6]
5. 2025Co-founds BioStack with Sanat Mishra. Mishra is chief executive and Patwa is chief technology officer.Reported by Indian Startup News and Bharat Fast.^[1]^[2]
6. 2026BioStack joins Y Combinator's Spring 2026 batch with a team of 11.Confirmed on both founders' Y Combinator profiles.^[2]^[4]
7. 2026The company announces a $9.5 million seed round on 22 October, led by Peak XV, DCVC and Afore Capital.The announcement and the trade press give the same investor list.^[1]^[3]

Who Parth Patwa is

Parth Patwa is the co-founder and chief technology officer of BioStack, a healthcare data company founded in 2025 and based in San Francisco. He runs the technical side of a business whose product is, in essence, a pipeline: hospital records go in, de-identified longitudinal datasets come out, and AI developers and biotech companies buy access. His co-founder Sanat Mishra, a cancer genomics researcher, is the chief executive.^[1][2]

Patwa's background is machine learning rather than medicine. He was a generative AI scientist at Amazon Web Services and, before that, a researcher at an MIT-linked public health organisation. His published work covers fact verification, multimodal misinformation, sentiment analysis of memes and language models for low resource Indian languages. That is a research record built on messy, multimodal, multilingual data, which is a fair description of hospital records as well.^[2][4][5]

He is the less visible of the two founders in the coverage of the funding round. Mishra gives the quotes in the company announcement and speaks about the patient record. Patwa appears in the photographs, is named as co-founder, and is credited with the technical architecture, but he has not given a detailed interview about the platform that this entry could find.^[2][3]

UCLA and the research training

Patwa holds a master's degree from the University of California, Los Angeles, according to his Y Combinator founder profile. UCLA's computer science department is one of the larger machine learning research groups in the United States, and a master's there is a standard route into industry research roles in California.^[4]

The profile does not say what he studied, when he graduated, or where he took his undergraduate degree. Those gaps matter for a biography because they are the usual anchors, and none of them has been filled from another source. What the publication record does show is that his research output began while he was working with academic collaborators, since the papers appear in conference proceedings and workshop tracks rather than only in company blogs.^[4][5]

The move from Los Angeles to industry research is the ordinary path for a machine learning graduate in that period. What is less ordinary is the subject matter he chose. Fact checking and meme sentiment are not the classic benchmark problems of computer vision or speech. They are problems about meaning, context and misinformation, and they require datasets that someone has to build before a model can be trained.^[5]

What his papers are actually about

Patwa's most cited indexed paper is a 2024 study on evidence backed fact checking using retrieval augmented generation and few shot in context learning with large language models. In plain terms, the system retrieves evidence and then asks a model to judge a claim against it, which is the architecture most fact checking tools now use.^[5]

A second cluster of work is about multimodal misinformation. Factify 2, published at a AAAI workshop in 2023, is a dataset of multimodal fake news and satire news. Memotion 2 and Memotion 3 are datasets for sentiment and emotion analysis of memes, including codemixed Hindi and English memes. FACTIFY3M, presented at EMNLP in 2023, is a benchmark for multimodal fact verification that asks models to explain their verdicts through the five W questions.^[5]

A third cluster is about languages that commercial models handle badly. Papers on parameter efficient fine tuning with synthetic data, hate speech detection in Devanagari script languages, and code mixed language modelling all point the same way: how to get useful performance out of a model when the labelled data is thin and the text switches between scripts mid sentence.^[5]

Recent work moves towards detecting synthetic media, with datasets for human versus AI generated image detection and human versus AI generated text detection posted in 2025 and 2026. The thread through all of it is data construction. Patwa's contribution in these papers is usually the dataset, the benchmark or the evaluation, which is the part of machine learning that medical AI buyers care about most.^[5]

From research to AWS

Before BioStack, Patwa worked as a generative AI scientist at Amazon Web Services. AWS published an author profile for him on a machine learning blog post about building AI driven policy creation for vehicle data collection and automation using Amazon Bedrock, which places him inside the company's applied AI organisation rather than in a pure research lab.^[4][6]

The AWS period is the bridge between academic publishing and a startup. At a cloud provider, a scientist works on products that customers deploy, with the constraints that come with them: latency, cost, privacy, and the fact that the data belongs to someone else. That is closer to BioStack's business than a conference paper is.^[4][6]

The Y Combinator profile also lists Vayu Robotics and Vimaan among previous employers, alongside Amazon and PathCheck Foundation. Neither company is described in the coverage of the funding round, and no dates are given, so the sequence of his career between UCLA and BioStack cannot be reconstructed with confidence.^[4]

PathCheck Foundation and the MIT connection

Indian Startup News and Bharat Fast describe Patwa as a former researcher at MIT. His Y Combinator profile lists PathCheck Foundation, with MIT in brackets, among his previous employers. PathCheck is a nonprofit that builds open source public health software and grew out of work at MIT, so the two descriptions are consistent: the MIT connection runs through the foundation rather than through a department appointment that has been documented.^[2][4]

No paper, project or date ties him to a specific MIT laboratory. The claim should therefore be read as an affiliation with an MIT-linked organisation, which is how his own profile presents it, and not as a faculty or postdoctoral position.^[4]

The public health connection is still relevant to BioStack. PathCheck works on software for health agencies and clinicians, and the problems it deals with, consent, record linkage, data minimisation, are the same problems a company selling de-identified hospital data has to solve in every market it enters.^[3][4]

Why a medical data company needs a CTO like him

BioStack's product is not a model. It is the infrastructure underneath one: workflows that pull clinical information out of separate hospital systems, link it to a patient over time, strip identifying details, and package the result as a dataset. That work is unglamorous and mostly engineering, which is why the technical co-founder matters more than the pitch.^[3]

The company says its platform includes workflows built around each provider's own data privacy protocols rather than a single global pipeline. In practice that means a separate configuration for every hospital group and every country, in four countries with different rules. A team of 11 cannot do that with a single script, and the design of the pipeline is the CTO's problem.^[1][3]

The data itself is multimodal. The company says it has combined several hundred thousand X-rays, MRI and CT scans with several million telehealth audio transcripts, plus pathology and other clinical records. Linking images, audio and text to the same patient timeline, and keeping the link after de-identification, is exactly the kind of dataset engineering Patwa's published work describes.^[3][5]

The seed round and the technical hiring

BioStack announced $9.5 million in seed funding on 22 October 2026. The round was led by Peak XV, DCVC and Afore Capital, with Afore Capital, DCVC, Emerson Collective, SV Angel, Verdict Capital, Streamlined Ventures and Y Combinator also participating. The valuation was not disclosed.^[1][3]

The company said the money would go into expansion across the United States and international markets, development of its technology platform, and hiring for engineering and research teams. For a CTO, that last item is the operative one: the round funds the people who build and maintain the pipeline.^[1][2]

The round is small by the standards of 2026 medical AI, where seed rounds for infrastructure companies often close above $3 million and growth rounds run into the hundreds of millions. Nine and a half million dollars buys a team of perhaps twenty engineers in San Francisco for a year, which sets the pace at which the company can add hospitals.^[1]

The citation claim, checked

The company and both news reports say Patwa has published more than 40 papers with over 1,800 citations at venues including CVPR, NeurIPS and EMNLP. His Y Combinator profile repeats the same figures. Those are self-reported numbers and they are the kind that grow with time and with the counting method.^[2][4]

Semantic Scholar, queried in November 2026, listed an author profile for Parth Patwa with 50 papers and about 1,480 citations. Google Scholar, which counts more generously and includes preprints and workshop papers, is the profile he links to himself. The order of magnitude is confirmed, the exact figures are not, and the difference between 1,480 and 1,800 is the difference between two databases rather than a contradiction.^[4][5]

The venue claim is partly supported. His indexed papers appear at EMNLP, LREC, COLING workshops, ICON and AAAI workshops such as DE-FACTIFY. CVPR and NeurIPS do not appear among his most cited indexed papers, so those two venues should be treated as claims rather than verified placements.^[5]

What has not been confirmed

No date of birth, age, family detail, marital status or height for Parth Patwa has been published in any source used for this entry, and none has been guessed. His undergraduate institution, the subject of his UCLA master's degree and the year he completed it are all unstated.^[2][4]

The career timeline is also incomplete. The Y Combinator profile lists Amazon, PathCheck Foundation, Vayu Robotics and Vimaan as previous employers but gives no dates and no titles beyond the AWS role. The description of him as a researcher at MIT is consistent with the PathCheck affiliation and is not evidence of an MIT appointment.^[4]

On the company, the hospital partners in the United States, India, Brazil and Argentina are unnamed, the valuation was not disclosed, and the Indian biotech partner in the cancer vaccine programme is unnamed. The claim of access to more than 100 million de-identified records is the company's own and has not been independently audited. The publication and citation counts differ between the company and Semantic Scholar, and both are recorded here rather than reconciled.^[1][2][3][5]