Subhabrata (Subho) Mukherjee

Also known as Subho Mukherjee, Subhabrata Mukherjee

Subhabrata (Subho) Mukherjee is an AI researcher whose work spans credibility analysis in online communities and language models for healthcare. He completed a computer science PhD summa cum laude at the Max Planck Institute for Informatics and was a 2018 SIGKDD dissertation runner-up. At Microsoft Research, he led foundation-model work that included Orca, a 13-billion-parameter model trained on GPT-4 explanation traces. His first-authored February 2026 paper used signals from 115 million patient-AI interactions and clinician-led testing. He is a co-founder and Chief Science Officer at Hippocratic AI, where he leads the technology organization building safe generative conversational AI for healthcare.

Now at: Hippocratic AI
Works: 126
Citations: 2,849
h-index: 28
Badges: Founder
Bridge

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

2017

A PhD on credibility analysis

He completes a computer science PhD summa cum laude at the Max Planck Institute for Informatics. His March 2017 thesis, “Probabilistic Graphical Models for Credibility Analysis in Evolving Online Communities,” builds Conditional Random Field models to assess the credibility of user-contributed content.
2018

The SIGKDD dissertation runner-up

He is named the 2018 SIGKDD dissertation runner-up.
June 2023

Orca at Microsoft Research

At Microsoft Research, he leads large-scale foundation-model efforts, including Prometheus and Orca, as a Principal Researcher. In June 2023, he is first author of Orca, a 13-billion-parameter model that learns from GPT-4 explanation traces and surpasses Vicuna-13B by more than 100% on Big-Bench Hard.
March 2024

The Polaris LLM constellation

In March 2024, he is first author of Polaris, a safety-focused LLM constellation for real-time patient-AI conversations. Its trillion-parameter system uses cooperating specialist agents, and evaluations include over 1,100 U.S.-licensed nurses and more than 130 U.S.-licensed physicians.
March 2025

Polaris 3.0 launches

In March 2025, Hippocratic AI launches Polaris 3.0, a 4.2-trillion-parameter constellation of 22 LLMs with a 99.38% clinical accuracy rate. He is a co-founder and Chief Scientist at the company.
December 2025

A NeurIPS health AI talk

In December 2025, he and Munjal Shah give an invited talk at the NeurIPS workshop The Second Workshop on GenAI for Health: Potential, Trust, and Policy Compliance.
February 2026

Production signals for clinical AI

In February 2026, he is first author of a paper that grounds conversational-AI safety in real-time signals from 115 million patient-AI interactions and clinician-led testing with over 7,000 licensed clinicians and 500,000 test calls.
As of October 2026

Chief Science Officer at Hippocratic AI

As of October 2026, he is a co-founder and Chief Science Officer at Hippocratic AI, where he leads the technology organization building safe generative conversational AI for healthcare. In 2026, the company launches Polaris 5.0, a 5-trillion-parameter constellation powered by a 700-billion-parameter core. Hippocratic AI has received $404 million in funding at a $3.5 billion valuation.