The Semantic Layer for Life-Science AI: Building Agents and Apps on the Amass API
About Event
Every biologist is about to have an agent. But the moment that agent needs to actually know something: the trials for a target, the papers behind an MoA, the regulatory precedent for an indication. Falling back to scraping PubMed, parsing ClinicalTrials.gov, and stitching identifiers together by hand is not good enough and where most life-science AI projects die.
Amass (https://amass.tech/) is the layer that removes it: a unified, cross-linked API over biomedical publications (BioMedCore), clinical trials (TrialCore, regulatory documents (RegulatoryCore), therapeutic molecules (DrugCore) and gene targets (GeneCore) with more in the works. All of it built for agents to consume, not for humans to click through. A paper knows which trials it references; a trial knows which papers describe it and much more.
This is a hands-on, demo-first session. We'll go from a fresh Amass API key to a working life-science agent live, and look at what it takes to build on top of Amass; whether you're wiring it into Claude, ChatGPT and friends via MCP, calling the REST API from a notebook, or shipping an app on the gallery.
Alexander will cover:
The shape of the problem — why "just add more sources" isn't the hard part, and why structure and cross-linking are
A live tour of the API: searching BioMedCore and TrialCore, fetching by Amass ID, and resolving external identifiers (PMID DOI NCT)
Plugging Amass into an agent via MCP — giving ChatGPT/Claude (or your own agent) a biomedical brain in a few lines
How tod use Amass with a coding agent such as Claude Code
Building an app on top: a worked example from query to UI, and what the gallery makes cheap
A reality check — what's accurate, what's token-efficient, what's still hard about building on us today, and where the roadmap goes next
Whether you're a bioinformatician tired of writing the same ingestion code, a CompBio researcher in academia, or building AI products in industry: if you want your tools to know biology instead of re-deriving it every time, this talk is for you. We'll keep ~20 minutes at the end for open Q&A and discussion.
About Alexander Junge
Alexander is Co-founder and CTO at Amass, where the team is building the semantic layer for life-science AI. He holds a PhD in Bioinformatics and worked in Data Science at Novo Nordisk and headed Machine Learning at the Danish healthtech scaleup Corti. Alex's research on building knowledge discovery tools for researchers received more than 18 thousand citations.