TL;DR
Tamarind Bio offers an inference platform that runs open-source molecular AI models (including AlphaFold) and packages them for non-technical scientists and developers. The company says it serves much of the top 20 pharma, dozens of biotechs and tens of thousands of scientists, and has added features like model fine-tuning, custom UIs for docker containers and wet-lab integrations.
What happened
Founders Deniz and Sherry launched Tamarind Bio as a centralized inference service for computational biology, born out of running large-scale model workflows in a Stanford lab. The platform aggregates leading open-source models used in drug discovery, provides a scientist-facing web app alongside programmatic APIs, and implements a standardized schema for model data formats. Tamarind built a custom scheduler and queue that horizontally scales workloads where individual inferences can take minutes to hours and run on single GPUs; jobs are split across CPUs and GPUs to optimize timing. As adoption grew among large pharma and biotech customers, the product expanded beyond a model library to include reproducible multi-model pipelines, support for onboarding customer models, fine-tuning, UIs for arbitrary docker containers, and connections to wet-lab data sources. The team is publicly hiring and invites feedback via their product and contact channels.
Why it matters
- Reduces dependence on technically specialized intermediaries in research groups by providing scientist-friendly tooling.
- Standardized model interfaces and pipelines can make computational experiments more reproducible and scalable across organizations.
- Built-in support for fine-tuning and private model onboarding addresses a common need as labs want to run proprietary workflows on shared platforms.
- Enterprises remain cautious about sending sensitive, un‑patented drug data to external providers; data security and trust are key adoption considerations.
Key facts
- Founders: Deniz and Sherry (announced on Hacker News).
- Tamarind is described as an inference provider for AI drug discovery and serves models like AlphaFold.
- Claims adoption: used by much of the top 20 pharma, dozens of biotechs and tens of thousands of scientists.
- Platform offers both a programmatic interface for developers and a scientist-friendly web application.
- Engineered a standardized schema to harmonize model data formats across tools.
- Built a custom scheduler/queue optimized for horizontal scaling; individual inference calls often take minutes to hours and run on one GPU at a time.
- Supports multi-model pipelines, reproducible protocols, model fine-tuning, UIs for arbitrary docker containers, and connections to wet-lab data sources.
- Product URL: https://app.tamarind.bio; company website: https://www.tamarind.bio; contact email listed as deniz[at]tamarind.bio.
What to watch next
- Whether large pharma will increase external platform use given concerns about leaking sensitive, un‑patented drug data.
- Adoption of Tamarind's fine-tuning, private model onboarding and wet-lab integrations among enterprise customers.
- Not confirmed in the source: details on Tamarind's funding, revenue, or specific customer contracts and names.
Quick glossary
- Inference provider: A service that runs trained machine learning models to generate predictions or outputs for user data, often offering APIs and infrastructure to scale those runs.
- AlphaFold: A deep learning model architecture used to predict protein structures from amino acid sequences; widely used in computational biology.
- Fine-tuning: The process of further training a pre-trained model on new or specialized data to adapt it to specific tasks or domains.
- Docker container: A packaged unit of software that bundles code and dependencies into a portable runtime environment.
- Scheduler/queue: Infrastructure components that manage job execution order and resource allocation across compute hardware like CPUs and GPUs.
Reader FAQ
What does Tamarind Bio provide?
An inference platform that runs open-source molecular AI models and packages them for both developers and non-technical scientists.
Who uses Tamarind?
The company says it is used by much of the top 20 pharma, dozens of biotechs and tens of thousands of scientists.
Can customers run their own models on the platform?
Yes — the company reports support for onboarding customer models, fine-tuning and building UIs for arbitrary docker containers.
Is Tamarind Bio publicly funded or acquired?
Not confirmed in the source.
How does the company handle long-running model jobs?
They built a custom scheduler and queue optimized for horizontal scaling, splitting workloads across CPUs and GPUs; individual inferences may take minutes to hours and typically run on one GPU at a time.
Hi HN, we're Deniz and Sherry from Tamarind Bio (https://www.tamarind.bio). Tamarind is an inference provider for AI drug discovery, serving models like AlphaFold. Biopharma companies use our library of leading…
Sources
- Launch HN: Tamarind Bio (YC W24) – AI Inference Provider for Drug Discovery
- Tamarind Bio is driving AI-assisted drug discovery with AWS
- Tamarind Bio: No-code bioinformatics for scientists
- Tamarind Bio
Related posts
- Locating Vehicle Photos in Seconds: GeoSpy’s SuperBolt Reaches 1-Meter Accuracy
- CES 2026 roundup — weird, wild and wonderful tech reveals from Las Vegas
- Kentucky rolls out mobile ID app; Apple Wallet support due this summer