Data privacy regulations such as GDDR, CCPA and HIPAA pose a challenge to train AI systems on sensitive data such as financial transactions., patient health records and user device logs. Historical data is what AI systems can “teach” to identify patterns and make predictions, but there are technical barriers to using it without compromising human identity.
One solution that has gained currency in recent years is federal tuition. The method By training a system on different devices or servers, the servers hold the data without exchanging data, allowing collaborators to build a common system without sharing data. Intel recently partnered with Penn Medicine to develop a brain tumor-classification system using federated learning, a group of major pharmaceutical companies including Novartis and Merck built a federated learning platform to accelerate drug discovery.
Tech giants, including Nvidia (via Clara), offer federal education as a service. But a new startup, DynamoFL, hopes to take graduates on a federal learning platform that focuses on performance, ostensibly without sacrificing privacy.
“DynamoFL was founded by two MIT electrical engineering and computer science PhDs, Christian Lau and myself; “We spent the last five years working on privacy-preserving machine learning and hardware for machine learning,” CEO Vaikunth Mugunthan told TechCrush in an email interview. After receiving repeated job offers from leading financial and technology companies trying to build federal education internally in light of privacy laws like GDPR and CCPA, we discovered a huge market for federal education. During this process it was clear that these organizations were struggling to support federated learning internally and we built DynamoFL to address this gap in the market.
DynamoFL – Saying that I have Key customers in the automotive, Internet of Things and financial sectors – the go-to-market strategy is in its early stages. (The startup currently has four employees, with plans to hire 10 by the end of the year.) But DynamoFL is focused on refining novel AI techniques to stand out from the competition, offering capabilities that dramatically increase system performance. fight Attacks and Vulnerabilities in Federal Education – Such as “member estimation” attacks that allow access to data used to train a system.
“Our personalized federated learning technology… Enable[s] Machine learning teams tune their models to improve performance on individual clusters. This gives C-suite executives more confidence when deploying machine learning models that were previously considered black box solutions. Muguntan said. “This [also] It sets us apart from competitors like Devron, Rhino Health, Owkin, NimbleEdge and FedML that struggle with traditional federated learning challenges.
DynamoFL promotes the platform as cost-effective relative to other privacy protection AI point solutions. SFederated learning doesn’t need to collect bulk data on a central server, DynamoFL can reduce data transfer and computing costs, Mugunthan asserts — for example, allowing the client to send smaller and more files than petabytes of raw data. As an added benefit, this can reduce the risk of data leakage by eliminating the need to store large amounts of data on a single server.
“Conventional privacy-enhancing technologies such as differential privacy and federated learning have been plagued by ‘privacy versus performance’ every year, where the use of stronger privacy-preserving techniques during model training inevitably results in poorer model accuracy. This critical bottleneck has prevented many machine learning teams from adopting the privacy-preserving machine learning technologies needed to protect user privacy while complying with regulatory frameworks, Mugunthan said. “DynamoFL’s personalized structured learning solution solves a critical barrier to machine learning adoption.”
Recently, DynamoFL closed a small seed round (4.15 million at a $35 million valuation) with participation from Y Combinator, Global Founders Capital and Basis Set. The startup is part of Y Combinator’s Winter 2022 batch. According to Muguntan, the proceeds will be primarily spent on recruiting product managers who will integrate DynamoFL’s technologies into future user-friendly products.
“The pandemic has highlighted the need to quickly use disparate data to respond to health care crises. In particular, the pandemic has emphasized how accessible critical medical information must be in times of crisis while protecting patient privacy. “We are well positioned to weather the technological slowdown. We currently have a runway of three to four years, and Tech slowdown has helped our recruiting efforts. The big tech companies were hiring most of the top federated learning scientists, so the big tech hiring slowdown created an opportunity for us to hire more federated learning and machine learning talent.