TurinTech has released Artemis calling it the “world’s first evolutionary AI platform.” It was announced at NVIDIA GTC 2025, along with the news that the company had raised $20 million in a Series A funding round led by IQ Capital and Oxford Capital.
The focus of Artemis is simple – reduce the AI-driven tech debt that organisations are beginning to incur. Given the amount of tech debt that most organisations struggle with, this launch is likely to attract a lot of attention. However, access to Artemis is by invite only and you need to apply here, to get on that program.
According to Dr. Leslie Kanthan, CEO and Co-founder of TurinTech, “AI-generated code is accelerating innovation, but it’s also accelerating risk—introducing inefficiencies, security vulnerabilities in a rising tide of unvalidated code. Enterprises need more than just AI-generated code; they need AI-evolved code which ensures performance, security, and cost efficiency.
“Artemis optimizes and validates any code for any application at enterprise scale, whether AI-generated, human-written code, or legacy code, so every line is secure, scalable, and production-ready. With strong demand from industry leaders, we’re excited to help shape the future of AI-driven software development.”
Why is AI creating technical debt?
Over the last two years, AI vendors have been pushing the idea that GenAI can resolve the significant code backlog that enterprises struggle with. It’s an attractive proposition, and the tools on offer range from code assistants to AI developers. According to software.com, developers can spend as little as two hours a day writing code. Compare that to an AI that can produce code 24 hours a day.
One of the claims for supporting AI-driven software development is that the AI will always produce syntactically correct code. That has an immediate impact on the time taken to write software. It removes problems when code is checked into repositories and takes away one significant level of coding error.
But, and it’s a hugely significant but, syntactically correct code does NOT guarantee that the code is secure or safe to use. For decades, developers have used sample code, free open-source code and sites like Stack Overflow. The code they get is correct but comes with a use-at-your-own-risk approach.
TurinTech also says that there is research showing that the code generated by GenAI can suffer up to 20% of hallucination rates. Those can introduce vulnerabilities and impact performance.
What is needed to deal with this?
To take advantage of AI-driven code, there is a need to write tests, examine the code, correct errors, retest and then release to production. Currently, most of these steps are performed by a human in the loop. Yet this brings other challenges. Do the developers and test teams know the language that the code is written in? Do they have access to the latest security threat data? Where they part of the design team for the app?
All of this adds workload to developers and test teams but, does it offset the time take to write code? According to another study quoted by TurinTech, “AI-assisted code reviews increased the average time to close pull requests by 42%, highlighting the growing burden of manually reviewing AI-generated code.”
To improve this, there has been a move towards AI-assisted testing. This is where the AI is tasked with defining tests to ensure the software is safe and secure. Usually, the same AI-engine is used to write the test as design the test. Just like asking developers to test and approve their own code, it does not make sense to allow AI to mark its own homework.
Enter TurinTech and Artemis
Recognising the problem, TurinTech set out to develop a tool that will take any code written by humans or AI. According to the announcement, it will then automate the codebase review, optimise and validate the code. All of this is done using agentic AI and is scalable to handle any size of codebase.
TurinTech states that this is significantly different from the existing use of GenAI for code review and testing. That is because “Artemis applies its proprietary Evolutionary AI to iteratively refine, validate, and evolve code – ensuring enterprise-grade performance, security, and scalability.”
Importantly, Artemis utilizes a combination of proprietary agentic AI and algorithms within its code review workflow. It also integrates with existing developer tools and can run on-premises or in the cloud. Both of these are key benefits that will appeal to potential customers.
The company has listed four key features for Artemis:
- Enterprise-Grade Code Optimization: Automatically refactors and optimizes any code—AI-generated, legacy, or developer-written—to improve performance, security, and maintainability
- Agentic AI Capabilities: Artemis combines agentic intelligence and evolutionary optimization to automate and improve critical tasks like code review, security validation and performance tuning, ensuring production-ready software
- Cloud & Hardware-Aware Optimization: Optimizes code for cloud, on-prem, and specialized hardware (e.g., GPUs, TPUs, and Edge devices) through real-time validation, benchmarking, and performance tuning
- Faster, Leaner, More Cost-Efficient Code: Artemis has been shown to optimize open-source code for libraries used in financial applications, improving runtime acceleration by 32.7%. In tests on open-source NVIDIA GPU libraries, Artemis uncovered and resolved critical defects in cuCollections and NCCL, addressing vulnerabilities that risked memory corruption, system crashes, and communication failures.
The company used Artemis to optimize OpenAI’s Whisper model. It achieved a 25% faster runtime while maintaining accuracy and reliability.
Enterprise Times: What does this mean?
Organisations cannot continue to take on more tech debt. Over the past two decades, every new technology shift has increased the level of tech debt they have to deal with. The size of the problem caused consulting firm McKinsey to develop a new way of measuring that debt. It also stated that CIOs believe tech debt now accounts for 20-40% of the value of their entire technology estate.
This announcement focuses on a solution that differs from what has been tried before. Artemis is not going to write code. It will help solve the problem of tech debt in code. As such, it does not mark its own homework and offers a fresh approach to the problem. Will it remove the need for a human in the loop? No. Will it make their life easier? On the face of it, absolutely, but only time will tell.
The announcement does raise one concern. The company states that it will accommodate any language. We’ve heard similar statements from vendors before. There is no list of languages on the TurinTech website, and that is something CIOs will want to see. After all, when you talk about being able to examine the entire enterprise codebase, you’d better have an extensive list of languages you understand.
