Today
Verta, the AI/ML model management and
operations company, formally launched with its product that helps data
science teams bring order to the chaos of sprawling enterprise machine learning
environments. Used by well-known global brands, including one of the
world's leading workplace collaboration tools, Verta helps enterprise data
science teams standardize otherwise fragmented workflows in order to ship
models more frequently with full traceability and low overhead.
"Data science teams live in their own specialized world, working with data,
running experiments and building great models. The software deployment teams
that take those models and use them to power production applications have
a completely different focus and set of trusted tools. Forcing these teams to
learn each other's tools is a distraction neither of them need. With
Verta, we help each team to stay focused on what they do best," said Manasi
Vartak, founder and CEO of Verta.
Vartak, who has worked at Facebook, Google, Microsoft and Twitter in software
engineering and machine learning roles, created ModelDB, the first open source
modern model management system, during her PhD at MIT's CSAIL lab. ModelDB
is widely used in the industry and Fortune 500 companies today and forms the
backbone of Verta.
Mike Leone, senior analyst at ESG, said: "As the AI/ML market expands rapidly,
data science teams are becoming overburdened. They're tasked with solving
strategic business problems, but they're bogged down by several non-data
science tasks: data integration, data quality, cumbersome checkpoints;
governance/compliance; learning intricacies of technology in the
operational stack, like Kubernetes. This is where Verta can help by taking on
the AI operational burdens data science teams face and enabling them to
concentrate on strategy, innovation and what they were hired for: data
science."
Verta interoperates with the rich variety of tools and workflows used by data
science and machine learning teams-including TensorFlow, PyTorch, Spark.ml and
R-allowing them to stay productive using the approaches they determine best
suit their needs. Verta's support for the rapidly changing landscape ensures
data science teams can continue to innovate quickly, rather than wasting
time supporting a brittle, home-grown patchwork of systems that need constant
care. Verta's MLOps capabilities have been designed to be compatible with
trusted application platforms such as Kubernetes, helping enterprise
infrastructure teams to support model-based applications
with well-established methods and tools.
"Before Verta, it used to take us about six months to deploy a new model into
production. This wasn't delivering value to customers fast enough. With Verta,
we can deploy models multiple times every month, so effectively ten times
faster, and with a lot less overhead for our team," said Jenn Flynn, senior
data scientist at LeadCrunch.ai. "Verta lets us focus on the data science
without worrying about infrastructure and operations."
For enterprises with large-scale model management needs, Verta's ModelDB-based
model catalog and governance mechanisms enable them to keep track of model designs,
deployment approvals and full traceability audits to provide the highest levels
of assurance demanded by the toughest global regulations. Customers can
design and deploy models with confidence, trusting that Verta's visibility
and reporting functions will show them what they need to know, when they need
it.
Verta's model monitoring capabilities help data science teams ensure their
models remain accurate and that their intelligent products keep providing value
to customers. By monitoring model performance, data drift and service
levels across deployment environments-including Verta's Inference Engine, AWS
SageMaker, and other domain-specific systems-Verta ensures model-based
applications continue to perform at their best, and can be rapidly refreshed as
and when needed. Data science teams can monitor the performance of their
models, and ensure they are well-informed of the state of production
applications at all times. IT teams also benefit from Verta's
integrations with standard monitoring tools, ensuring they see the views
they need in the formats they prefer.
Today Verta also announced $10 million in Series A funding led by Intel Capital
with participation from General Catalyst who led the seed round.
Mark Rostick, VP and senior managing director at Intel Capital, said: "Verta is
addressing one of the key challenges companies face when adopting AI - bridging
the gap between data scientists and developers to accelerate the
deployment of machine learning models. Companies need a solution that solves
the DevOps part, deploys ML models into production, monitors the ML model
performance and accuracy, applies governance and supports reproducibility.
Verta connects data scientists, DevOps and production engineers and
enables them to more quickly and easily create and deploy efficient ML
solutions that give their companies a competitive edge."
Steve Herrod, Managing Director at General Catalyst, said: "AI/ML is such a hot
area, but there are so few companies focusing on the challenges faced by organizations running
intelligent products in production. Developing theoretical models is one thing
but turning them into products that deliver real value to customers is quite
another. I'm excited by Verta's focus on the part of the market with all
the value."