Drug Discovery AI
P

Persist AI

Persist AI sells formulation software and a robotic formulation lab to pharmaceutical and biotech development teams. Persist AI Suite, which it calls a self learning AI formulation scientist, runs inside the customer's own environment on AWS, GCP, Azure or premises, with a private knowledge index built from the customer's reports, assay and stability tables and legacy design of experiments records. Specialized agents combine Bayesian optimization, deep neural networks with transfer learning, density functional theory and molecular dynamics to flag API, excipient, stability and solubility risks before experiments begin, propose formulation strategies, forecast shelf life, predict in vivo behavior through PBPK modeling and design DOE protocols, returning a report within 24 hours. New lab results feed back to refine the recommendations, and the installed engine sends no molecule data out.

Cloud Lab lets customers design experiments in a browser for Persist's robotic lab in West Sacramento to build and test, across amorphous solid dispersions, SEDDS, microspheres, depots, injectables and topicals, with results on a dashboard and customer IP kept out of shared models. Persist also runs staffed formulation projects and custom automation. Karthik Raman is chief executive, the company raised a $12 million Series A in May 2025, it collaborates with Nivagen on long acting injectables, and its scientists published an automated microsphere screening study with AstraZeneca researchers in 2026.

AI Health Index verifiedNovember 21, 2026
Compare Persist AI with other vendors
Founded
—
Headquarters
West Sacramento, California
Website
www.persist.ai/
Categories
drug-discovery
Indexed Products
Persist AI Suite, Cloud Lab
Buyer Segments
Pharmaceutical formulation scientists, Biotechnology companies, Pharmaceutical development organizations
Assessment

Capability Axes

An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read

AI Capability
AA on AI CentralityThe artificial intelligence is the product. Remove the model and there is nothing left to sell.
Vendor Published

The software is the model. Persist AI Suite exists to predict: it screens a molecule's formulation risks, proposes strategies, forecasts stability and in vivo behavior and designs experiments, and its value is the recommendation a scientist would otherwise reach through months of bench work. The private knowledge index only stores what the models read, so taking the models out leaves the customer with a document store and nothing else to buy. Persist's robotic lab and staffed projects are separate offers, and the lab's structured data is what trains the general models the suite starts from.

BB on Autonomy and Oversight ModelThe oversight structure is described and one part is missing, commonly the threshold at which the system stops or what happens after it is wrong.
Vendor Published

Scientists decide what to test. The suite is described as augmenting formulation staff rather than replacing them: reports separate findings grounded in literature or the customer's data from computational predictions that need experimental proof, show a confidence figure for lead candidates, and wait on the team, which feeds new results back to refine later recommendations. In Cloud Lab the customer designs each experiment and reads the results in the dashboard. No threshold is set at which a prediction is withheld or marked too uncertain to act on, and no path is described for rejecting or correcting a recommendation.

BB on Model and Technology TransparencyThe approach or the suppliers are named without the version and update discipline behind them. Naming a supplier is the entry to this band both here and on Model Supply Chain Disclosure, which ask different questions of the same disclosure: who receives the data, and what produces the output.
Vendor Published

The method is named. Specialized agents combine Bayesian optimization, deep neural networks with transfer learning, density functional theory and molecular dynamics, built to work from little data. The base models draw on formulation literature, public pharmacology, peer reviewed research, patents and regulatory filings plus thousands of formulation records, and Persist's robotic lab supplies structured data tied to process parameters.

Each customer gets its own model and private index, fed with its PDFs, spreadsheets and legacy design of experiments records. No model versions or update practice are published, so a team cannot tell when the engine it runs has changed.

BB on Model Supply Chain DisclosureSubstantial partial disclosure, or a chain that is structurally short: an in house build, a cleared model that cannot be quietly swapped, or a deployment where the transfer does not occur at all. Naming the model provider exits the band below into this one; the axis rises from here on the completeness of the party list and on the terms that govern data once it arrives.
Vendor Published

The chain is short for the installed suite. The engine and the customer's model and index run on the customer's AWS, GCP, Azure or premises infrastructure, the product page promises zero outbound data for the molecule, and Persist says it has no access to the customer's private data.

The same FAQ also offers a deployment hosted in Persist's own cloud, where that promise cannot hold in the same way, and Cloud Lab results live on Persist's web service, though customer IP is kept out of shared model training. No subprocessor list is published for either hosted route.

CC on Clinical and Operational EvidenceNamed customers, or vendor reported percentages with no method, denominator or reference standard. Scale of use is recorded here and is not treated as evidence of benefit.
Vendor Published

Nivagen is the named collaborator, working with Persist on an AI driven manufacturing process for a long acting injectable, with no results reported. Persist's August 2026 paper in Pharmaceuticals, written with AstraZeneca formulation scientists, screened 129 IgG microsphere formulations on the robotic platform and linked additive class and antibody concentration to early release, using classical statistics with no learned model, so it measures the lab rather than the suite's recommendations.

A 2025 company post reports one client reaching its target formulation in 28 days after six months elsewhere, and data generated 30 times faster than by hand, with no method behind either. No evaluation of the suite's predictions against experiments is published.

BB on AI Safety and PHI StewardshipCategorical commitments are published, such as no training on customer data, without the retention schedule behind them.
Vendor Published

Molecule data stays inside the customer's perimeter for the installed suite, Persist says it cannot access the customer's private data, and customer IP from Cloud Lab is kept out of shared model training, with formulation IP and commercialization rights left with the customer under the MSA. The data protection addendum provides for return or deletion of personal data when the agreement ends, subject to legal retention. No retention schedule is given for hosted experimental records or reports, and no procedure covers a recommendation that proves harmful or wrong.

Regulatory and Compliance
CC on HIPAA and BAA PostureA status is stated but not supported: compliance is claimed, or business associate status stated, without the underlying document, including where the only privacy notice published covers the website rather than the service that handles patients. The mirror case grades here as well: a substantive privacy document that reaches the service, with no statement of business associate status anywhere and no scope position taken. One of the two elements the band above asks for is present, which is a grade below it and not a grade at the floor.
Vendor Published

Formulation work rarely touches patient data, and the suite is built around drug, excipient and experiment records. When Persist handles personal information under its data protection addendum, the customer is controller and Persist processor, with terms on use, security, transfers and deletion at the end of the agreement. Business associate status is not stated, and nothing says whether clinical data, such as human pharmacokinetic inputs to PBPK modeling, may be loaded or how it would be protected.

DD on Security Certifications and Trust CenterControls are asserted with nothing independent behind them, or nothing is published. Read the note before concluding anything: this is the grade most often corrected on a second pass, because assurance material frequently sits on a parent domain or inside an old announcement rather than on the product pages.
Vendor Published

No certification is listed. The installed suite inherits the customer's own infrastructure controls, and Cloud Lab work runs under standard pharmaceutical confidentiality agreements. For personal data, the addendum requires safeguards with regular assessment of their effectiveness, lets a contract customer request an inspection or use an independent auditor under set conditions, and obliges Persist to mitigate security incidents. No SOC 2 report, ISO 27001 certificate or penetration test covers the suite, the hosted option or the lab portal.

BB on FDA and Regulatory StatusThe pathway is stated and in progress, or a clearance is named without the vintage and scope a buyer needs to match it to the product on offer. A product outside device regulation also sits here when the vendor publishes what it is for and in what setting and the product on offer matches, or names another regime it works under, such as a registration outside the United States. A general clause saying the service is not medical advice does not reach this band.
Vendor Published

Research and development of drug formulations is the stated use. Reports organize API, excipient, stability, toxicology and process risks for a formulation team, separating evidence based findings from predictions meant to be tested, and Cloud Lab runs screening experiments on the team's designs. The suite is sold to formulation and development scientists, who stay responsible for experimental validation and regulatory filings. Nothing in it reaches a patient, and the product on offer matches that research purpose.

DD on AI Governance and Bias DisclosureNothing published on how model behavior is governed or tested. Multilingual operation with no subgroup performance sits here when the vendor markets recognition quality as a strength, because a caller the system failed to understand leaves no complaint and no record.
Vendor Published

Reports separate evidence based findings from predictions and show confidence for candidates, and lab results feed back so a team can see where a prediction missed. None of that evaluates the models. No results are reported across dosage forms, molecule classes or data volumes, though the suite is pitched as ready for small data, and no policy or outside review governs how the customer specific models are trained and changed.

CC on AI Liability and RecourseMechanisms exist that let someone challenge an output, such as audit trails, source traceability or review before commit, with nothing standing behind the output and no route for the harmed party.
Vendor Published

Formulation reports mark which findings rest on literature or data and which are predictions, so a scientist can weigh a candidate before testing it, and the customer can save project snapshots and attach later results. Under the MSA customers keep formulation IP and commercialization rights with no royalties. The data protection addendum requires Persist to mitigate security incidents involving personal data. A recommended candidate that fails in the lab or later in development carries no guarantee, correction commitment or remedy.

Integration and Deployment
DD on EHR and Interoperability DepthNo integration evidence. A connector described as available on request grades here until one exists.
Vendor Published

Formulation data, not patient records, moves through the product. The suite ingests PDFs, scanned reports, assay and stability spreadsheets and legacy design of experiments files into its index and emails reports back, Cloud Lab returns results to a dashboard, and custom automation projects build instrument interfaces. No electronic health record or clinical data standard is involved, and no integration with a laboratory information system or electronic lab notebook is named.

BB on Deployment Model and Data ResidencyOptions and residency are stated with isolation or the processing path left open.
Vendor Published

Deployment options are named: AWS, GCP, Azure or on premises for the suite, which lives inside the customer's infrastructure with zero outbound molecule data, a style hosted in Persist's own cloud listed in the FAQ, and Cloud Lab as a separate web service over the West Sacramento lab. The data protection addendum commits AI processing of personal data handled by Persist to California. No storage location or isolation is stated for the Persist hosted option or for Cloud Lab records and backups.

Commercial
CC on Commercial TransparencyNo price is published and the posture is discoverable: a buyer can establish how the product is sold and what drives the cost before contacting the vendor. Most of the index sits here.
Vendor Published

Pricing is custom, set by model tier, data volume and deployment style, and quoted through a contact form, with one complimentary assessment per organization. Cloud Lab, staffed formulation projects and custom automation are priced separately, and the Cloud Lab catalog lists 29 standard studies. No figure is published for any suite tier or lab study, though the MSA leaves formulation IP with the customer with no royalties or fees.

CC on Setting and Specialty CoverageCoverage is claimed broadly without specifics, or stated clearly with nothing validating it yet.
Vendor Published

Dosage forms are listed in detail: amorphous solid dispersions, SEDDS and lipid systems, long acting injectables, microspheres, in situ depots, creams and topicals, parenterals, oral solids, nanoparticles and extended release platforms, with process simulation for coating, lyophilization, spray drying, hot melt extrusion, tableting and microencapsulation. The buyers are formulation and development scientists at pharmaceutical and biotech companies.

The microsphere screening study tested the robotic lab rather than the suite, and the Nivagen injectable work has no results yet, so no dosage form has validation of the suite's recommendations behind it.

Alternatives

Alternatives to Persist AI

Other products that do the same kind of work as Persist AI, and what each one does.

BenchSci develops EMET, an AI research environment for scientists in biopharma and academia. The platform combines scientific literature, experimental databases and internal research with a biomedical knowledge graph.

Variational AI develops Enki, a generative model for designing small molecules against a research team's target product profile.

Medra builds a Physical AI Scientist for preclinical drug research: robotic lab workcells and scientific AI that design, run, interpret and improve experiments in a closed loop.

Aqemia designs small molecule drug candidates with Qemi, a platform that joins physics based prediction of how tightly a molecule binds its target with generative AI that proposes new molecules.

XtalPi is a drug and materials discovery company founded in Boston in 2015 by three MIT postdoctoral physicists, headquartered in Shenzhen and listed on the Hong Kong Stock Exchange (2228) since 2024.

Molcure designs antibodies and peptides for drug companies, using a large language model for antibodies that it built from scratch together with physics based simulations and experimental data from its own biolab.

Commercial

Pricing

Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.

Entry Price Pricing Basis BAA Tier Implementation Source
Custom quote
Persist AI Suite is priced to each customer by model tier, data volume and deployment style (the customer's AWS, GCP or Azure account, on premises, or Persist hosted), with one complimentary assessment per organization. Cloud Lab access, staffed formulation projects and custom automation are separate offers. The data protection addendum covers personal information processed by Persist. No business associate agreement is available to review. Installation and implementation scope depend on the customer's environment and project. Vendor Published

The license rate and the units used to bill for the customer deployed software are agreed with Persist. Hosted laboratory work and staffed formulation projects have their own commercial terms.