Inflexa Cloud Platform

Inflexa Cloud runs AI analysis agents for computational biology on AWS. A scientist describes the question and the cohort in plain language. The agent reads the literature for that biology, proposes an analysis plan, waits for approval, then runs every step in a sandbox and writes out results, figures and a report. Each step is recorded as it happens: the method, the parameters, the input files, the software versions and the model that made each decision.

Most biopharma teams already run AI agents on research data, usually on a frontier model through the vendor's API. Inflexa Cloud keeps that workflow and changes the layer underneath it. The reasons to switch are cost, grounding and where the data lives.

Cost. The Knowledge Plane is a curated, versioned body of computational biology method knowledge: rules about which method fits which situation, parameters with their sources, tested script templates and the alternatives that were rejected. It does the deterministic part of planning, so the model does less. In our September 2026 planning benchmark, an open model with the Knowledge Plane scored higher than the frontier model on its own, at 13 to 19 cents on the dollar per run. Teams that were paying for frontier reasoning on every step can move most of that work to lower-cost models and spend the difference on compute.

Grounding. Every method choice in a plan cites the rule behind it. Literature claims come with PubMed identifiers. The agent queries more than 30 biological databases live during the run, including Ensembl, UniProt, Open Targets, ChEMBL, Reactome, KEGG, STRING and GEO, instead of recalling them from training. In the same benchmark, plans produced with the Knowledge Plane contained zero fabricated references.

Provenance and lineage. The record is written by the execution layer. The model never authors it. Each output carries a content hash, its inputs, the method and parameters, and the model that reasoned about it, aligned with W3C PROV. The record is signed, so a later edit is detectable, and it can be verified offline with an open-source tool. A reviewer can ask any figure where it came from and walk the lineage back to the raw data.

Sandboxed execution. Analysis code runs in a hardened container with R, Python and the bioconda tools already installed. The container runs as a non-root user with all Linux capabilities dropped and outbound network access denied unless it is allowed for a specific step.

Your data stays yours. Connect your own S3 bucket. Data is pulled in for processing and removed from Inflexa's side once results are written back to your storage. Privacy Mode, on by default for Biotech and Enterprise plans, means no prompt or response is retained beyond serving the request and nothing is used to train a model. Enterprise customers can deploy Inflexa Cloud inside their own AWS account, so processing and storage stay in their VPC and compute is billed by AWS directly.

Model choice. The architecture treats the model as replaceable. Use Inflexa-managed model access, or bring your own API key on Biotech and Enterprise and pay your provider directly. Whichever model runs a step, its resolved name is in the record.

Teams work in shared projects with role-based access, audit log export, and SAML single sign-on with SCIM on Enterprise. Long analyses run for days or weeks and survive restarts. Scientists steer through the plan: force a covariate, reject a method, ask for a second one, approve before anything runs, and sign off on the report.

Typical programs: target validation, biomarker discovery, patient stratification, responder analysis, mechanism of action, cross-study meta-analysis and toxicogenomics.

Inflexa Cloud is built on the open-source Inflexa orchestrator (Apache 2.0). Analyses are files in open formats with a provenance document you can verify without us. If you stop using Cloud, the open-source tool runs everything you produced on your own machine.

Infrastructure is hosted on AWS. Data is encrypted in transit with TLS 1.2 or higher and at rest with AES-256. SOC 2 Type II, ISO 27001 and GDPR alignment are in progress. HIPAA BAA and 21 CFR Part 11 support are available on Enterprise. Security documentation is available on request from security@inflexa.ai .