What challenges did Tangram Therapeutics face before AI implementation?

Long timelines and inefficiencies in traditional drug discovery. Need to accelerate and scale drug target discovery and evaluation for RNAi medicine development.

What was the impact of the AI solution?

Document processing that previously took weeks now takes hours. Tangram can assess four to five target-indication propositions in a few hours instead of one quarter, improving speed by up to 50x. The platform processes about 300x more data than before, with roughly 288 million input tokens and 16 million output tokens per day. LLibra OS enabled researchers to explore questions and hidden targets that were previously out of reach.

Source

ExploringEvidence: Medium65/100

Tangram transforms drug discovery with agentic multi-LLM AI on AWS

Use case typeDrug discoveryUpdated Jun 13, 2026

Tangram Therapeutics is a UK biotech company focused on solving human disease through computational RNA interference (RNAi) medicines. To accelerate drug target discovery and evaluation, Tangram built LLibra OS, an agentic AI platform that unifies proprietary, licensed, and curated public datasets for research and target-indication assessment. The platform supports retrieval augmented generation, web search, and text-to-SQL to help researchers identify novel targets, evaluate therapeutic potential, and design medicines.

Organization: Tangram Therapeutics
Industry: Pharma
Location: United Kingdom
Published: June 2026

Reported outcomes

300x

data processed daily increaseOther quantified impact

50xtarget-indication evaluation speedup288 million tokensdaily input tokens16 million tokensdaily output tokens

Strategic outcomes

Speed & agilityAccelerated target discovery and evaluationScale & capacityUnified large biological datasetsNew product / capabilityEnabled novel target explorationNew product / capabilitySupported target and medicine design

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Target-indication evaluation speedup: 50 x increase

AWS Customer StoriesJun 8, 2026Customer storyExplicit claimMedium evidence strength

Increasing speeds by up to 50 times.

Normalized claim

Data processed daily increase: 300 x increase

AWS Customer StoriesJun 8, 2026Customer storyExplicit claimMedium evidence strength

LLibra OS processes 300 times more data.

Normalized claim

Daily input tokens: 288 million tokens increase

AWS Customer StoriesJun 8, 2026Customer storyExplicit claimMedium evidence strength

roughly 288 million input tokens ... on a typical day

Normalized claim

Daily output tokens: 16 million tokens increase

AWS Customer StoriesJun 8, 2026Customer storyExplicit claimMedium evidence strength

and 16 million output tokens on a typical day

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Tangram Therapeutics
Provider: AWS
Maturity: Exploring
Linked source: AWS Customer Stories

To accelerate drug target discovery and evaluation, Tangram built LLibra OS, an agentic AI platform that unifies proprietary, licensed, and curated public datasets for research and target-indication assessment

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1Drug discovery
2Research and development
3Knowledge discovery

Long timelines and inefficiencies in traditional drug discovery.
Need to accelerate and scale drug target discovery and evaluation for RNAi medicine development.

Built LLibra OS as an agentic AI platform on AWS.
Used Amazon Bedrock for generative AI and agentic orchestration.
Used AWS Glue for data ingestion, preparation, and integration across more than 1,000 biological datasets.
Used Amazon Athena for interactive query and analytics.
Added retrieval augmented generation, web search, text-to-SQL, and an in-house evaluation harness to keep models and components current.

Technologies

Amazon Bedrock AWS Glue Amazon Athena RAG text-to-SQL

Document processing that previously took weeks now takes hours.
Tangram can assess four to five target-indication propositions in a few hours instead of one quarter, improving speed by up to 50x.
The platform processes about 300x more data than before, with roughly 288 million input tokens and 16 million output tokens per day.
LLibra OS enabled researchers to explore questions and hidden targets that were previously out of reach.

Architecture

LLibra OS runs on a serverless AWS architecture and uses Amazon Bedrock for access to and evaluation of LLMs, AWS Glue for discovering, preparing, and integrating data, and Amazon Athena for interactive analytics. The system has an agentic orchestration layer, retrieval augmented generation, web search, text-to-SQL, modular LLM swapping, and an in-house evaluation engine.

Sources & evidence1

Evidence: Medium65/100Evidence strength