Last Friday, Arlan Rakhmetzhanov (17), a high school student from Almaty, Kazakhstan, and his team at Nozomio Labs, publicly launched Nia, the latest AI co-pilo
Last Friday, Arlan Rakhmetzhanov (17), a high school student from Almaty, Kazakhstan, and his team at Nozomio Labs, publicly launched Nia, the latest AI co-pilot for software developer teams.
Rakhmetzhanov, who works as a researcher for Stanford University’s venture capital initiative, founded Nozomio Labs a year ago. His organization creates both educational courses on technology and developer tools, like their new project, Nia.
Currently in beta, Nia will soon become a low-cost AI agent focused specifically on enterprise and engineering teams. The software acts as a collaborative “teammate”—a developer co-pilot for engineering teams that helps with the understanding of codebases.
“Nia supports both private and public repositories. Users can add a project, and the agent essentially scans a whole repository, no matter the size,” said Rakhmetzhanov. “We provide a bunch of integrations with GitHub, GitLab, Slack.”
Because of Nia’s efficiency for developers working on open-source software projects, the service ranked third globally on internet company Product Hunt’s daily ranking of “Top Products Launched.” Nia’s immediate popularity even caused Nozomio’s servers to temporarily crash on Friday as users simultaneously indexed repositories with up to 70 million tokens, per Rakhmetzhanov.
This new technical software stands out from other AI tools that use vector embeds and multi-query retrieval to explain codebases. Nia can scan entire codebases at a magnitude that OpenAI’s ChatGPT is currently unable to.
“ChatGPT has only a 128k context window, which is very small for large repositories. So you can’t just pass the whole thing into ChatGPT,” said Rakhmetzhanov. “Nia stores everything in the vector as vector embeddings. It converts every word, every letter, gets the number, stores it into the database, and makes a retrieval.”
“Recently we indexed the whole HuggingFace repository, which is more than one gigabyte and 14 million tokens. That is equivalent to 30 novels, or the whole trilogy of Lord of the Rings 30 times. It took just 15 minutes, and the accuracy was 99.7 percent,” said Rakhmetzhanov.
Nia is ideal for collaborative software engineering teams with developers constantly joining codebases and adding to programming projects. It can be accessed through team networks, such as the popular work platform Slack. Nia will provide new software engineers with a full breakdown of the project when they begin work.
“Imagine you have a team of engineers, and you’re onboarding them. It’s easy to ask questions to Nia about your repository,” said Rakhmetzhanov. “How do we use front-end here? How do we use front-end there?”
When creating Nia, the Nozomio team used the framework of other AI agents to produce a tool that may not transform the software engineering process entirely but certainly makes collaborative project work more efficient.
“I don’t think we’re going to ‘change’ the game as much as just the efficiency of the players which will—as a consequence—allow for developer teams and technology enthusiasts to change the game on their own,” said Chief Operating Officer Finn Järvi.
In addition, Rakhmetzhanov and his team have reached over 4,600 students with their ten-course curriculum since starting Nozomio. Rakhmetzhanov spent two months creating the first educational course.
“It’s basically AI from zero to one. You make an AI app from scratch, even if you’re a beginner, so that anyone can learn,” said Rakhmetzhanov.
According to the website, Nozomio students can “learn by building” through modern web development and an introduction to generative AI. The platform also offers hackathon competitions and bite-sized video lessons.
Rakhmetzhanov’s personal coding education began last year when he took AP Computer Science at his high school. Bored with simple coding and Python coursework, he took up more ambitious coding projects on his own.
“When I was 10, my dad started telling me about how programming and AI will change the world. He actually envisioned that AI would boom in 2023-2024 when I was only 10, seven years ago,” said Rakhmetzhanov. “I wasn’t that interested in coding in general; I just liked business as a whole. But then my dad told me about Y-Combinator, and I just got hooked by the tech community and how everything is done.”
Rakhmetzhanov’s interest in coding and commerce has only spiked since then. In 2023, he published a 50-page book on the basics of venture capital, entitled Doorway to Venture Capital: An Introduction for Young Students. In Jan. 2024, Rakhmetzhanov launched Hanlan Tech and began making iOS and web applications, the biggest of which, “Extracurrify”, generated 85,000 page visits.
In the fall and winter of 2023, Rakhmetzhanov picked up real-world experience as a software engineer intern for KAZ Minerals, a copper mining company. There, he developed a tool “designed to identify and simplify the mining and copper extraction processes” and improved the company’s AI model.
Over the winter, Rakhmetzhanov met Washington, D.C. student Finn Järvi through a fellowship in the Technology and Entrepreneurship Center at Harvard University. Järvi, also 17 years of age, now serves as Chief Operating Officer on the founding team of Nozomio Labs and Nia AI. Despite the impressive long-term results, Rakhmetzhanov’s and Järvi’s journey to launching Nia was not as smooth as it may seem.
“Before we launched Nia, we would spam every corner of Reddit and online coding forms so that people would schedule a demo with us. For every 300 people I reached out to, 5 responded. 1 booked. And repeat. Contacting laboratory directors at Stanford or professors for their personal input really helped us improve our product,” said Järvi.
Eventually, Nozomio picked up traction and Rakhmetzhanov found a mentor in a tech founder from the Y-Combinator Summer 2023 batch, a cohort of the very startup accelerator Rakhmetzhanov was inspired by as a child; this founder has used Nia as a demo in his own work. Additionally, Keywords AI founder Hendrix Liu from YC W23 posted online last Friday on LinkedIn, promoting Nia’s launch.
Since Oct. 2024, Rakhmetzhanov has also worked as a research assistant for Stanford University’s venture capital initiative. Through the project, the 17-year-old uses Python scripts to replace antiquated websites for famous Silicon Valley firms like Sequoia Capital, Menlo Ventures, and Founders Fund.
As Rakhmetzhanov continues to advance his expertise in machine learning and software development, he will implement his knowledge into Nozomio’s course offerings. He will also use his experience of the for-profit world and venture capital space to generate profits through Nia. The Nia agent will be a software subscription service, but the whole educational side of Nozomio will remain free, assures Rakhmetzhanov.
“Nozomio education is completely free because I believe that education should be free and should be accessible for everyone, especially in the age of AI,” he said.
Currently, Rakhmetzhanov is at work on his for-profit business model for Nia. Rakhmetzhanov plans to charge 15 cents for every API request and a 20-dollar premium subscription for the service.
In the future, Rakhmetzhanov hopes to expand upon Nia and create a fully-automated and autonomous AI software engineer. He sees Artificial Intelligence as a powerful tool to apply to all areas of life, work, and academia.
“I use AI everyday, and I would really love a tool that deeply understands the whole codebase and doesn’t just give you responses to help you code but can actually execute tasks on its own,” said Rakhmetzhanov. “So we’re building towards agentic AI.”
“The future, in my opinion, is where you don’t have to hire but can just build your own AI tools. That’s what happened during the .com bubble, and that’s what’s happening right now. It applies to all industries, from health tech to developer tools to legal stuff,” said Rakhmetzhanov.
