#VBTransform of @AnnaGriffinNow @jeggers @manuaero @may_habib @mmarshall @nickfrosst @parasnis @PhilipDawson @sharongoldman @stevewoodwho @uljansharka @Venturebeat
Head over to our on-demand library to view sessions from VB Transform 2023. Register Here
AI is like a baby with the memory of an elephant: It’s never going to forget anything, said Amruta Moktali, chief product officer for Skyflow.
“We can’t forget that all the data we are using to train these models need to be protected,” she told the crowd at VentureBeat Transform 2023.
>>Follow all our VentureBeat Transform 2023 coverage<<
Skyflow, the winner for Best Presentation Style in the Innovation Showcase at VentureBeat Transform 2023, is built on top of an enterprise’s data vault to help secure sensitive data. The company’s platform helps to isolate, protect and govern data privacy, which is critical when it comes to training generative AI and large language models (LLMs), according to Moktali.
VB Transform 2023 On-Demand
Did you miss a session from VB Transform 2023? Register to access the on-demand library for all of our featured sessions.
Introducing Skyflow Vault
The company has added Skyflow Vault for LLMs to its existing privacy platform that features proprietary polymorphic encryption and fine-grained access controls, Moktali explained.
The API-based platform isolates and protects sensitive data during modeling. It features embedded access controls, and users can securely connect to third parties and perform encrypted searches on sensitive data.
Skyflow de-identifies sensitive data, helping to ensure compliance with regulations while not compromising the end user experience.
Supporting multiple use cases
For instance, the company worked with a healthcare company to aggregate unstructured doctors’ notes, import them into a HIPAA environment, de-identify and certify them, then turn them back for model training and continual fine-tuning.
In other use cases, Moktali pointed out, organizations may have sensitive terms that are not necessarily personally identifiable information (PII) — for example, sensitive product names. Those can be imported into the Skyflow UI and sensitive data directory to ensure they are redacted.
“We’re able to cover multiple use cases,” said Moktali, emphasizing that the platform is model-agnostic.
“You can use whatever model you want,” she said. The protection of data fed into models is Skyflow’s primary concern. “You want to make sure that it doesn’t contain any sensitive information.”
She also noted that models will evolve, and when data is protected, enterprises gain the freedom to experiment.
“Different use cases need different models,” she said. “You can’t restrict yourself to one or the other.”
The Innovation Showcase highlighted 10 companies in the generative AI, machine learning (ML) and analytics spaces. The three winners were Skyflow, Unstructured.io (Most Likely to Succeed) and Arize (Best Technology).
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.