Monday, November 16, 2026

Nota Inc. Targets Financial Sector AI Coding Market; GPU Throughput Up by Up to 95%

Input
2026-11-16 10:50:18
Updated
2026-11-16 10:50:18
Courtesy of Nota Inc.
[Financial News] Nota Inc., a company specializing in AI inference optimization, is entering the AI coding agent business with a product that can be used even in the closed-network environments of the financial sector. The system runs AI on a company’s internal servers without sending source code to an external cloud. In particular, Nota is focused on easing the infrastructure investment burden of AI adoption for financial firms by optimizing AI models and inference environments to increase by up to 95% the number of requests that can be processed on the same graphics processing unit (GPU). The move is expected to help expand the use of on-premises AI in the financial sector, where security regulations and high infrastructure costs have constrained adoption.
According to industry sources on the 16th, Nota Inc. signed a contract with Dube, a financial information technology (IT) company, to supply optimized large language models (LLMs) and servers for an on-premises AI coding agent tailored to the financial sector, and has completed delivery.
Under the contract, Nota Inc. provided an integrated package tailored to Dube’s AI coding agent, covering everything from AI model compression and inference and serving environment optimization to server configuration.
Dube is a company that carries out systems integration (SI) development projects for financial institutions. Its AI coding agent supports a range of development tasks, from analyzing and organizing development requirements to developing screens and interfaces. It can also incorporate coding rules and internal libraries tailored to a client’s development environment.
The financial sector is constrained in adopting cloud-based AI coding tools because it uses closed-network environments that restrict the external transmission of source code and internal business information.
The on-premises AI coding agent supplied by Nota Inc. and Dube runs LLMs on clients’ internal servers. It can perform AI-based development tasks without calling the APIs of external AI services, allowing sensitive development information to be managed internally.
Taking into account the nature of coding agents, which repeatedly process large volumes of code and development guidelines, Nota Inc. applied model compression and inference optimization technologies.
A coding agent may receive tens of thousands of tokens in a single request, so GPU memory usage and response times can increase as the number of concurrent users grows. Nota Inc. used quantization to reduce LLM memory requirements and optimized both the answer-generation process and the serving environment for handling multiple requests.
As a result, with the same GPU configuration, the number of requests processed per minute increased by approximately 58%–95% compared with before optimization. The time from submitting a question to generating the first answer was also cut to half or less. This means that more users’ requests can be handled with the same hardware. For businesses, it provides a way to reduce the burden of adding more GPUs even as the number of AI coding agent users grows.
Nota Inc. plans to work with Dube to promote the adoption of on-premises AI coding agents among financial-sector clients and expand the supply of optimized LLMs and servers tailored to clients’ user numbers and security and operating environments.
[email protected] Jang Min-kwon Reporter