Physician-Engineer | Google Research | Creator of LangExtract
New York, NY
- Tweets 179
- Following 113
- Followers 357
- Likes 294
Pinned Tweet
Excited to share work from @SRSchmidgall @valentinlievin and our teams, on how RL can improve clinical performance of a base Gemini model even in highly nuanced domains like psychiatry! The improvements are outstanding and have broad implications.
We trained Gemini via RL to conduct psychiatric intake with simulated patients.
In a double-blind study with 32 psychiatrists playing the role of patients, we show that multi-turn RL in simulation improves Gemini’s performance on structured text-based psychiatric intake. 🧵
2
8
524
Akshay Goel retweeted
Trust in clinical AI cannot be benchmarked into existence. It must be earned through prospective studies in real-world clinical settings.
Today, @Google and @BIDMChealth published the first clinical study of AMIE, our research medical AI, in @TheLancet! This prospective study is a huge step on our multi-year journey toward enabling universal access to expert-level care.
Two of my biggest takeaways:
💯100% conversationally safe: Zero safety stops were required across all encounters. This gives us the confidence to continue expanding our US study with Included Health.
🤝 Stronger doctor-patient relationships: 75% of PCPs found AMIE’s pre-visit outputs helpful, shifting limited consultation time away from rote information gathering and toward collaborative care.
We're just getting started!!
🔗 AMIE Overview: goo.gle/amie
📄 Lancet Study: goo.gle/4i7Fx9N
✍️ Google Blog: goo.gle/amie-blog
📰 NYT coverage: nytimes.com/2026/10/08/techn…
Huge congrats to the team, especially @HardyShakerman @apalepu13 @AdamRodmanMD @PeterBrodeurMD!!
Today @TheLancet published our AMIE study with @BIDMC_Medicine, the first prospective study of patient-facing conversational diagnostics in a real clinic.
AMIE is a research system people can chat with before doctors appointments. Testing AMIE outside the lab with real urgent care patients showed how technology can help gather medical histories safely. Clinicians reported that the summaries helped them prepare for visits in 75% of cases and influenced their approach to care in more than half. AMIE’s differential diagnoses also matched the doctors’ final diagnoses 90% of the time giving clinicians more time to focus on patient care.
It’s Google’s first-ever paper in The Lancet’s main journal!
3
2
15
1,056
Akshay Goel retweeted
Our teams @Google + @BIDMC_Medicine publish world's first prospective study of a patient-facing conversational diagnostic intelligence @TheLancet - Go #AMIE team!!! blog.google/innovation-and-a…
Today @TheLancet published our AMIE study with @BIDMC_Medicine, the first prospective study of patient-facing conversational diagnostics in a real clinic.
AMIE is a research system people can chat with before doctors appointments. Testing AMIE outside the lab with real urgent care patients showed how technology can help gather medical histories safely. Clinicians reported that the summaries helped them prepare for visits in 75% of cases and influenced their approach to care in more than half. AMIE’s differential diagnoses also matched the doctors’ final diagnoses 90% of the time giving clinicians more time to focus on patient care.
It’s Google’s first-ever paper in The Lancet’s main journal!
5
7
53
2,504
Exciting work by @SRSchmidgall, @taotu831, and colleagues on how Gemini accelerates scientific discovery! 🙌
Today, we're excited to share early progress on using Gemini to accelerate scientific discovery in the real-world. We present an extension of Co-Scientist which we use to collaborate with scientists across materials science, biology, and computer science.
5
452
Very nice new interactive visualization of #LangExtract by @HuggingApps :)
Btw: the library was released over a year back but the HuggingFace live demo was recently added.
Google just dropped LangExtract on Hugging Face Spaces
An interactive demo for their new library that extracts structured information from unstructured text using LLMs with precise grounding and interactive visualization
▶️ huggingface.co/spaces/google…
1
162
Akshay Goel retweeted
Incredible advance in AI doctor development! A new Google model learns medicine from simulated clinical environments through a reinforcement learning approach called ResidencyRL! This real world modeling was the missing link in turning AI fully into a specialist clinician!
Doctors don't become doctors from textbooks. They become doctors through residency, requiring years of practice across thousands of patient encounters.
Can AI learn the same way?
Introducing ResidencyRL 🧵
6
35
191
19,882
Akshay Goel retweeted
1/ Medical students don't become doctors by memorizing facts and passing exams. They do by residency, years of practice across thousands of patient encounters.
We wanted to see if we could train models the same way.
Introducing ResidencyRL 🧵
4
9
48
10,996
Akshay Goel retweeted
Thrilled to share our new preprint on AMIE, demonstrating expert-level performance compared to primary care physicians in real-time clinical video consultations!
4
14
77
12,858
Exciting work from our team published today! 🚀
How far can RL improve frontier models on agentic, real-world diagnostic assessment? Spoiler: a lot!
Incredible collaborative effort. Stay tuned for more updates and a huge shoutout to my co-authors!
arxiv.org/abs/2608.07418
Medical training begins in textbooks, but clinical competence is forged through practice. We introduce ResidencyRL: a multi-turn online RL method to train health agents in simulation, and demonstrate significant gains on top of frontier systems. arxiv.org/abs/2608.07418
ALT Paper visual abstract
1
3
14
1,183
So great chatting with @jmorgan and learning from his experience building @ollama. It’s a core part of the #LangExtract stack, so this was especially fun!
4
159
Very cool application of #LangExtract!
Looking to Build GraphRAG? Start with This Practical Guide from our star, Pavan
In this practical guide, Pavan demonstrates how to build a GraphRAG architecture using LangExtract, @neo4j, Qdrant, and @ollama.
You’ll learn how to:
→ Extract entities and relationships from unstructured text with LangExtract
→ Build a knowledge graph in @neo4j
→ Use Qdrant for semantic retrieval
→ Combine graph traversal with vector search for more context-aware responses
→ Run the entire pipeline locally with @ollama
If you’re exploring GraphRAG or building production-ready retrieval systems, this is a great walkthrough.
Read the article:
blog.stackademic.com/a-pract…
3
224
Nice Python list here for 2026! 🙌 #langextract
🐍 10 Python Libraries You Shouldn’t Ignore in 2026
1️⃣ Polars - Ultra fast DataFrame library built in Rust ⚡
2️⃣ MarkItDown - Convert documents into Markdown for LLM apps 📄
3️⃣ Smolagents - Lightweight framework for AI agents 🤖
4️⃣ PydanticAI - Build reliable GenAI apps with structured outputs 🧠
5️⃣ FastMCP - Easy way to build MCP servers for AI tools 🔌
6️⃣ LangExtract - Extract structured data from messy text 📊
7️⃣ GPT Pilot - AI coding assistant for development 👨💻
8️⃣ Pyrefly - Fast modern static type checker 🔍
9️⃣ Data Formulator - Generate charts using plain English 📈
🔟 Mostly AI - Create privacy safe synthetic datasets 🔐
2
124
Akshay Goel retweeted
Thrilled to share that our paper on AMIE—our conversational medical AI—is now published in @Nature. We wanted to tackle one of the most challenging areas in clinical AI: management reasoning over multiple patient visits.
nature.com/articles/s41586-0…
9
56
202
22,539
Foundational work by @valentinlievin, @apalepu13, and @HardyShakerman on one of the earliest robust demonstrations of agents for longitudinal care management!
Our paper “Towards Conversational AI for Disease Management” is out in @Nature! We ran a blinded study comparing our multi-agent AMIE system to 21 primary care physicians to test longitudinal patient care management.
🔗 nature.com/articles/s41586-0…
@GoogleDeepMind @GoogleResearch
1
102
Amazing work here by colleagues @apalepu13 @hardyshakerman and the teams at Google + BIDMC!!
Today we announce results from a first-of-its-kind study with @BIDMC_Medicine on AMIE, our conversational AI for clinical reasoning. In a real-world clinical study, AMIE was found to be safe, feasible, and well-received by patients.
Learn more: goo.gle/4sXCogz
2
153
Excited to share we're launching a nationwide study with @IncludedHealth, showing how #Gemini (optimized on rigorous eval) can be safely and effectively deployed in real-world clinical environments.
With: @GoogleResearch, @GoogleDeepMind & P&D.
Blog: research.google/blog/collabo…
Today, @Google is announcing a strategic partnership with Included Health on a nationwide randomized study of AI in virtual care!
Official announcement: research.google/blog/collabo…
This is all possible because of the wonderful partnership across Included Health @GoogleResearch @GoogleDeepMind @GoogleForHealth and Google Platforms & Devices!
1
8
586
Excited for #NeurIPS2025 & San Diego weather!
Heading to the conference? Stop by @Google's booth at #1533 — @GoogleResearch @GoogleDeepMind
I'm showcasing #LangExtract & Gemini for Structured Radiology Reporting @ Kiosk 2 on Thurs, Dec 4 (1-5 PM).
research.google/conferences-…
1
180
Akshay Goel retweeted
I’m really excited about our release of Gemini 3 today, the result of hard work by many, many people in the Gemini team and all across Google! 🎊
We’ve built many exciting new product experiences with it, as you’ll see today and in the coming weeks and months.
You can find it today on @GeminiApp and AI Mode in Search. For developers, you can build with it now in @GoogleAIStudio and Vertex AI.
blog.google/products/gemini/…
The model performs quite well on a wide range of benchmarks.
203
334
3,394
400,051
Great video on #LangExtract from @Sam_Witteveen — Loved the demo at the end :)
Thanks for making this Sam!
piped.video/watch?v=t-53fouK…
1
3
817
Akshay Goel retweeted
For anyone who is interested -- I ported this to typescript and added an ability to use OpenAI not just Gemini.
npmjs.com/package/langextrac…
1
3
6
479