Qodo helps developers banish bad code with time-saving AI tools

Tech leads and their teams are actively exploring the potential of artificial intelligence (AI) as a means of boosting productivity and efficiency. AI generated code has the potential to streamline developer workflows and help them focus on higher priority tasks, but there is another key part of the puzzle to consider—code quality.

Giving hardworking developers time back means nothing if they then have to invest those hours into correcting errors in generated code—it’s simply swapping one problem for another. That’s where AI startup Qodo can help, equipping development teams with AI-powered agents and tools to help them review, test, and rectify coding errors as quickly—and accurately—as possible.

Qodo collaborated with Amazon Web Services (AWS) and NVIDIA through the AWS Generative AI Accelerator to accelerate model training and product deployment.

Building the solution to a trillion-dollar problem

Qodo was founded in 2022 by CEO Itamar Friedman, and Chief Product Officer Dedy Kredo, Qodo as an AI software platform that could help developers build high-quality software with confidence, precision, and ease. The Qodo leadership team combines deep technical knowledge and business acumen with previous experience founding successful startups like Visualead, and working at leading enterprises like VMWare and Alibaba Group.

Today, Qodo has established itself as a leader in AI innovation. In 2024, the company was named the Audience Choice winner at the AWS Unicorn Tank and recently raised $40 million USD in Series A funding. The Qodo team are now working towards a future where quality is embedded in every aspect of software development, from code inception to deployment and beyond.

“Qodo is a quality-first coding platform that enables professional, busy teams to code, review, and test complex software,” says Friedman. “Bad quality code is a trillion-dollar problem. You can write a lot of lines of code with AI, but you cannot automatically create Fortune 5000 software. At Qodo, we believe that if we solve code testing, we solve software.”

As the amount of AI generated codes increases so does the risk of mistakes. If AI code is to succeed, development teams need efficient tools to help them review, correct errors, and ensure code works the way the developer intended. Qodo’s platform fills that role, focusing on automated testing and code review rather than just code completion. But it’s not as simple as asking one AI to check another AI’s work.

Coding isn’t always an exact science. Professional developers can be opinionated about what ‘good’ looks like, and when there’s multiple ways to solve      a problem, the obvious fix may not be as obvious as it may at first seem. To tackle this problem, Qodo’s agents dynamically learn best practices of the company using it to collect the right context for a user’s request, before leveraging that context to surface problems with greater precision. For even deeper context-awareness, enterprises can index their full codebases with Qodo’s advanced methods for remote retrieval augmented generation (RAG), powered by the company’s advanced code embedding model.

“Companies are starting to understand that their real bottlenecks are around code review and code testing,” says Friedman. That realization has been key to Qodo’s success. “Our code-embedding model is unique and the best out there for solving complex codebases. We have one million developers installing our tools, one million pull requests being reviewed every quarter, [and] 50,000 tests being generated every day.”

Accelerating model training and deployment

Being an AI startup means operating in a hyper-competitive and fast-changing landscape. Success relies on more than a great idea and willingness to put the work in. Access to advanced technologies, tools, expert guidance, and a supportive community are all critical to powering growth. That’s why Qodo took part in the Generative AI Accelerator, a global 10-week hybrid program designed to help startups prove what’s possible in their industry.

Startups that successfully enrolled in the program could access up to $1 million USD in AWS Promotional Credits to build, train, test, and launch their products using the full breadth of AWS services and technologies. They learned from a curated and seasoned network of experts from AWS, presenting partners like NVIDIA, and other industry leaders from top AI companies.

All participants receive dedicated guidance from executive, go-to-market, and technical mentors matched to their industries and requirements. They can also connect with fellow cohort members from around the world through a growing virtual community, and attend exclusive networking events with AWS executives, AWS Partners, and top tier investors. For Qodo, the Generative AI Accelerator program presented a valuable opportunity to work closely with AWS and NVIDIA.

Delivering a better experience to a wider audience

To succeed and deliver the best possible experience for its customers, Qodo needs to cater to a rich set of deployment scenarios. After completing the Generative AI Accelerator, Qodo can confidently—and quickly—deploy its products across a wide range of infrastructure.      

“Our clients are Fortune 500 companies, and they want the Qodo platform to run inside their own Virtual Private Cloud (VPC), that often means running on AWS,” says Friedman. For example, some Qodo customers are only willing to use AI solutions that are compatible with Amazon Bedrock, a fully managed service providing access to high-performing foundation models (FMs) from leading providers. “Fortunately, AWS makes it very easy for us to deploy and run our models the way our customers want,” says Friedman. AWS customers can also now access Qodo’s products in AWS Marketplace, delivering an even more streamlined deployment experience.

AWS Marketplace serves millions of AWS customers and helps software companies expand the reach and accessibility of their products. For startups like Qodo, AWS Marketplace helps to deliver a more flexible experience for customers, with support multiple pricing options including pay-as-you-go, long-term commitment plans, Private Offers, and flexible payment schedules. Joining AWS Marketplace has also opened new co-marketing and co-selling opportunities for Qodo.

Leading technology and deep integration

Qodo is using AWS and the NVIDIA accelerated computing platform to improve the training and compatibility of its products. For example, training AI models is a resource-intensive process that requires significant time and financial investment. The Qodo team are using NVIDIA Hopper graphics processing units (GPUs) to train their models efficiently, delivering high-memory bandwidth, optimized processing power, and faster deep learning.

“At Qodo, we have four different models, and we use NVIDIA H100 to train them,” says Friedman. “We care a lot about quantization and the Hopper architecture in NVIDIA H100 enabled us to achieve 10x context length and 3x throughput.”  This helps the Qodo team avoid bottlenecks and increase the computation power available to them throughout training processes.

Beyond training, making its models compatible with major code repository systems is key to how Qodo delivers exceptional experiences for customers. Qodo Merge, a generative AI-powered assistant that complements the traditional code review process, features deep integration with Amazon Bedrock’s advanced generative AI models. This enables Qodo Merge to analyze code efficiently, generate high-quality suggestions, and provide automated pull request descriptions and walkthroughs. The result? Developers work more productively, and the cognitive load placed on reviewers can be alleviated.

The next generation of code generation

Going forward, Qodo is continuing to develop new ways to support developers with AI. “The second phase of Qodo is moving from a platform with a focus on code review and code testing to building the AI developer for complex code,” says Friedman. “I think we will see developers taking the software world into areas we could not imagine, and Qodo is the milestone, a critical steppingstone towards the next generation of code generation.”

AWS and NVIDIA will continue to help Qodo innovate and grow through a combination of cutting-edge technology and deep expertise. The company recently announced Qodo-Embed-1, a new state-of-the-art model for code embedding. Embed-1 is available in the Amazon Bedrock Marketplace and includes deep integration with NVIDIA NIM microservices. “NVIDIA and AWS are great partners with great programs, like NVIDIA Inception and the Generative AI Accelerator,” says Friedman. “They open opportunities for us. They understand our needs, what we're trying to achieve right now, and they help us be a better startup.”

If you want access to funding, expertise, and cutting-edge technology that can help your startup grow, consider applying for the Generative AI Accelerator. You can find the full details here.