ChordexBio brings Candor, a Filipino-built AI analyst for cancer mutations, to the Google Cloud AI Builder Cup 2026

Cancer genomics produces more data than most teams can interpret. A single tumor can carry hundreds of somatic mutations, and making sense of them has traditionally required specialized bioinformatics skills, dedicated compute, and time that many hospitals, laboratories, and biotech teams do not have.

ChordexBio is taking that problem to the Google Cloud AI Builder Cup 2026, a JAPAC-wide hackathon run by Hack2skill and sponsored by Google Cloud. Our entry, Candor, competes in the Future of Work & Enterprise Productivity challenge — and as a Filipino team applying AI to biology, we are bringing a different kind of everyday work to the stage: interpreting biological data.

Meet Candor

Candor is an AI-powered cancer mutation intelligence framework developed by ChordexBio to bridge the gap between complex biological data and informed decision-making. It was trained on more than 2.5 million mutations from The Cancer Genome Atlas (TCGA) — 11,160 patients across 33 cancer types — and its engine draws on Covary, ChordexBio's machine-learning framework for whole-genome analysis.

For the AI Builder Cup, Candor is embedded in Gemini as an agentic AI. A user pastes or uploads a tumor's somatic mutations. Gemini understands the request and calls Candor, Candor profiles the mutations against TCGA reference cohorts, and Gemini explains what the evidence does — and does not — support.

Four things Candor does

1. Turns a mutation list into a profile. Paste or upload a table, check it in a live validation preview, and get calls for cancer type, disease stage, and patient sex — each with scores and a confidence classification.
2. Shows its evidence. Every call comes with supporting markers and the closest reference profiles, so users can see why Candor says what it says.
3. Says "not sure" when it should. Weak or ambiguous evidence is reported as indeterminate or shared, not overstated — and Candor asks for genomic coordinates instead of guessing them.
4. Remembers the conversation. Follow-up questions build on earlier results, and chats can be exported or printed to share with colleagues.

Built on Google Cloud

Candor uses Gemini on Vertex AI, with function calling, as its reasoning agent. The web app runs on Cloud Run, and Candor's proprietary engine stays private as a separate Cloud Run service secured with IAM and Secret Manager. Gemini decides when to run an analysis; the mutation data is passed to Candor exactly as entered — the language model never retypes it.

How Candor changes the future of work

Today, a question about a mutation profile often becomes a request in someone's queue: reformat the data, run the pipeline, interpret the output, write it up. For institutions with limited compute and data-analytics capacity, that queue can be the difference between acting on genomic knowledge and not having access to it at all.

Candor moves that work into a conversation. By combining specialized computational analysis with agentic AI, it aims to reduce the time needed to access, interpret, and synthesize mutational and clinical knowledge, and to turn complex information into decision-support outputs. Specialists spend more time on judgment and less on reformatting, and teams without a dedicated analytics group can take part. The approach is designed to extend beyond oncology to medicine, biotechnology, healthcare, and the life sciences.

Responsible by design

Candor is a research prototype, not a diagnostic device. Its calls come from matching against public reference cohorts and must be reviewed by a qualified clinician. Candor does not offer treatment advice, and our demonstration uses public-cohort data only — never real patient data.

Watch Candor in action

See how a pasted mutation table becomes an evidence-backed profile, and how Candor handles uncertainty, in our demonstration video.

▶  Watch Candor in action

We will share more as the AI Builder Cup progresses. ChordexBio is open to conversations with research groups, institutions, and healthcare and biotech organizations interested in piloting Candor or applying the approach to their own data. For technical details or to arrange a demonstration, contact ChordexBio or visit our contact page.