Back to research overview

Research program

AI for Adjuvant Design

A four-stage effort to build the infrastructure, models, verification systems, and agents needed for trustworthy AI-assisted adjuvant research.

Four-stage visual roadmap progressing from adjuvant research infrastructure to a domain model, neuro-symbolic verification, and an autonomous research agent
One connected program: structure the field, build domain intelligence, verify scientific claims, and close the loop with autonomous agents.

Roadmap

Four connected stages

01Foundation

Benchmarks & Infrastructure

Formal task definitions, structured domain knowledge, datasets, and evaluation protocols for adjuvant research.

View benchmark
02Model building

Adjuvant Domain Model

A multi-agent synthetic-data pipeline and domain-specific fine-tuning for stronger adjuvant knowledge and reasoning.

03Verification

Neuro-Symbolic Verifier

SAVANT combines literature evidence, immune outcomes, and symbolic mechanism chains to inspect generated designs.

04Agentic research

Auto Adjuvant

An adjuvant research agent that connects evidence retrieval, candidate design, verification, and iterative planning.

Direction

Toward a verifiable design loop

The goal is not merely to generate candidates, but to make every step inspectable, evidence-grounded, and useful to domain researchers.

View publications Discuss collaboration