AI Infrastructure for Life Science

Tree-of-life–scale genomics search built on object storage. Fast, accurate and 10× cheaper.

Open console Sign in

Research

Inspect nuclear pore complexes, proteins, pockets, variants, annotations, confidence, and structural neighbors through a research surface built for fast biological questions.

npc evidence

query, measure, export

pdb:7r5k

structure

ready

Nup98 FG

region

42 residues

chain A:214

confidence

plddt 38.4

pocket npc-03

surface

18 residues

128

rows

refs

guesses

Researchers get a structure-native loop for looking at a nuclear pore, pulling measurements, and deciding the next query.

Features
Nuclear pore evidence
Residue inspection
Pocket measurements
Variant mapping
Structure neighbors

Search

Serverless search index over the largest biological datasets.

Search index

One serverless index surface spanning biological datasets.

live map

GenomesProteinsAntibodiesVirusesStructuresLigands

Genomes

assemblies, genes, regions

Proteins

sequences, folds, annotations

Antibodies

chains, CDRs, binding evidence

Viruses

hosts, strains, proteomes

Structures

PDB, AlphaFold, domains

Ligands

neighbors, pockets, contacts

Platform teams can find sequence homologs, structure neighbors, domain/profile hits, and evidence rows over immense corpora through serverless indexes designed for a better recall, latency, and cost curve.

Features
Serverless workers
Object cache
Parquet tables
ClickHouse serving
Provenance

Agents

Rafflesia exposes a biological toolbox as fast and scalable agent-callable primitives, letting agents resolve targets, fetch structures, run searches, inspect pockets, query tables, and keep provenance in the loop.

Terminal

Claude Codev2.1.3Opus 4.7

> use the biological toolbox to answer the next evidence question

*Resolve APP first so the target, organism, and source identifiers are explicit.

>rafflesia refs resolve"APP human" --json

|uniprot:P05067 - amyloid beta precursor protein - Homo sapiens

*Fetch the AlphaFold structure and keep the object id visible for downstream calls.

>rafflesia proteins structure fetchuniprot:P05067 --source alphafold --json

|okAF-P05067-F1 - object sha256:b639... - cache_hit true

*Before asking a structural question, inspect confidence so low-quality regions stay explicit.

>rafflesia proteins confidence summaryAF-P05067-F1 --json

|770 residues - median pLDDT 86.4 - low_confidence 94 residues

*Run pocket detection with explicit parameters and preserve the tool snapshot.

>rafflesia proteins pockets detectAF-P05067-F1 --probe-radius 1.4 --json

|ok8 pockets - largest 612 A^3 - tool fpocket 4.2 - warnings []

*Materialize residue features as a table so the agent can run a smaller follow-up query.

>rafflesia proteins residues tableAF-P05067-F1 --output residues.parquet

|ok770 rows - parquet object sha256:81e2... - provenance included

*Check ligand neighbors next, but keep source limitations in the answer instead of guessing.

>rafflesia proteins ligands neighborsAF-P05067-F1 --radius 4 --json

|warn0 ligands - warning: AlphaFold model has no bound ligand records

*That warning is useful. Fall back to a fold-search hit with experimental ligand context.

>rafflesia proteins folds searchAF-P05067-F1 --database pdb --limit 5 --json

|ok5 hits - top 6WXM_A - tm_score 0.92 - object sha256:9a14...

*Now the next query can be precise: compare pocket residues against the experimental hit.

>rafflesia proteins ligands neighbors6WXM_A --radius 4 --json

|ok12 neighbors - ligand HEM - chain A - provenance included

*Evidence chain ready: target, structure, confidence, pockets, tables, warnings, and object refs.

Agents can move from a small biological question to typed evidence, inspect the result, adjust parameters, and launch the next query without leaving their workflow.

Features

Enterprise

Run biological evidence infrastructure inside the boundaries pharma and biotech teams need: private data, controlled execution, audit logs, and deployment flexibility.

Enterprise teams keep sensitive structures, assays, annotations, and tool runs under their own storage, network, and access controls.

Features
Private storage
VPC deploy
Access controls
Audit trails
Self-hosting

Latest

Current Rafflesia surfaces for serverless biological infrastructure.

Serverless biology runtimeOpen the Rafflesia console

Search

Query serverless search indexes over the largest biological datasets.

Agents

Expose small, typed, reproducible calls for query-reflect-query loops.

Start building biological agents on Rafflesia

For scientific teams. Open the console to inspect serverless biology primitives, object ids, cache state, table artifacts, warnings, and provenance.

Open console Sign in

For developers. Use the API and CLI as thin wrappers over deterministic primitives that agents can compose.

Health