TurinTech is a leading provider of Loading...artificial intelligence optimization, enabling businesses to build and deploy accurate AI models to automate the end-to end Loading...data science lifecycle. The London-based startup was launched by four UCL doctoral graduates in 2018 to tackle the inefficiency of manual Loading...machine learning development and coding processes within their professions.
“These were very time consuming, resource intensive, and required deep domain expertise,” says Leslie Kanthan CEO and Co-founder of TurinTech. “To overcome those difficulties, we decided to convert our research into a product.” That led to the development of evoML, an AI optimization technology that enables businesses to rapidly build and deploy models for a variety of use cases.
TurinTech is a strategic partner of Exasol, leveraging our in-memory analytics database to help its customers generate high performing models within our fast database in order to support decision making and drive better business outcomes.
How TurinTech drives AI optimization
It’s crucial for the most ambitious organizations to optimize their AI infrastructures, but here’s a need to manage increasingly advanced databases storing ever larger volumes of data. Successful model development requires grappling with this scale complexity, and it’s becoming more difficult to solely rely on human input.
Enter evoML – TurinTech’s AI platform designed to generate highly efficient, explainable, and advanced models at faster speeds. “Our product generates scalable AI, with very high accuracy and efficiency. evoML speeds up the entire end-to-end data science process from months to weeks, and from weeks to days.”
Kanthan says it achieves this by optimizing model efficiency for quicker inference speed, lower memory use, plus a reduction in energy consumption to drive sustainability.
The platform also provides visual explanations and cues to streamline AI work for non-technical staff, potentially allowing them to run projects independently, freeing up specialist ML engineers for other tasks, and overseeing multiple projects to provide a boost in productivity. Model code is also provided to assist technical staff in their work.
Pioneering sustainable AI
By streamlining model deployment, evoML enables more sustainable AI development. Training AI is an energy intensive process, with the average application footprint equivalent to five times the emissions of an average car. But evoML can reduce these carbon emissions by 50% according to TurinTech. The method behind this “green AI” includes optimizing model code efficiency for quicker inference speed, combining reduced memory and energy consumption with multi-objective optimization to accelerate development and reduce emissions.
These features have allowed the evoML platform to generate an unprecedented number of models for numerous clients, with over 100,000 developed across 5,000 datasets last year.
“Large tech companies are using a huge amount of energy and that’s taxing for businesses and the economy,” says Kanthan. “We’re seeing that’s a big area, moving into sustainability and making sure that energy is pushed down to a minimal – what’s necessarily required. This is a significant economic benefit for companies as a whole. By reducing your energy footprint, you’re also reducing your costs.”
TurinTech and Exasol
Exasol has entered into a strategic partnership with TurinTech, combining our high-performance database with its AI optimization capabilities to help customers.
As organizations grapple with larger quantities of information, processes are being made difficult given the prevalence of ‘dirty data’, that’s siloed, mislabelled and unstructured. This further problematizes the training process and often produces low-quality results. Kanthan says that access to the right database technology is important in order to navigate this.
“Having well sourced, consistent and labelled data is key to successful AI transformation. In an increasingly data-driven world, organizations need to process information to make optimal, informed decisions in real time. This requires faster data infrastructures, and faster AI. That’s why it’s important to have the right database in place that centralizes and processes information from multiple sources.”
Towards an intelligent future
The pandemic has lead to increased digitization across a range of industries, with companies realizing the value of AI optimization. Leslie says it’ll be important for organizations to adapt in order to maximize value, otherwise they’ll be left behind. “Missing out on this will prevent access to new marketplaces, and there’ll be stiff competition from other companies entering that space. You’ll also risk being outdone by other tech giants looking to muscle in, so it’s about being AI-ready, and then moving on to be AI efficient.”
That means organizations need to transform their architectures by using the latest AI tools, in conjunction with advanced analytical databases, so they can develop the models they need to stay ahead.