Ippu Senkin Co., Ltd. (Head Office: Chuo-ku, Tokyo; Representative Director and CEO: Shuya Suzuki; hereinafter “Ippu Senkin”) has launched “IS Code” and “Local-AI Agent,” autonomous AI agents equipped with advanced reasoning and task-execution capabilities, built on NVIDIA AI infrastructure.

By fully harnessing the exceptional computing power of NVIDIA accelerated computing, this solution goes beyond conventional chatbots to deliver an agentic AI system that can think autonomously and complete complex business processes end to end. In a fully isolated on-premises environment, it provides secure and high-performance AI agent capabilities for industries with stringent security requirements, including financial institutions.

Background and Development Purpose

Many companies are currently adopting generative AI, and demand is rapidly growing for AI that can go beyond summarization and response generation to actually execute business operations. At the same time, cloud-based generative AI services present challenges such as the risk of external leakage of customer and confidential information, while usage-based pricing can create high ongoing operating costs that hinder sustained adoption. In particular, full-scale use of generative AI in financial institutions has been constrained by strict information management regulations.

To address these issues, Ippu Senkin, with support from NVIDIA, leveraged state-of-the-art hardware and LLMs to develop AI agents that deliver high-speed, high-accuracy inference in on-premises environments. This cloud-independent architecture strongly accelerates digital transformation for enterprises that require secure operating environments.

 Key Features of Ippu Senkin’s On-Premises AI Agents

Ippu Senkin provides autonomous AI agents that run on on-premises servers equipped with NVIDIAGPUs.
From document creation integrated with internal databases to coding tasks, the agents autonomously process a wide range of work in local environments. By adopting GPUs based on the NVIDIA Blackwell architecture and supporting multiple models including NVIDIA Nemotron, the solution achieves high-speed, high-accuracy operation.
Because it runs entirely on-premises, it combines low cost with strong security. Building on this advantage, Ippu Senkin aims to expand beyond financial institutions to a broad range of industries, particularly manufacturing and infrastructure, both in Japan and internationally.

The solution is available in two forms: “IS Code,” a CLI (command-line interface) for engineers, and “Local-AI Agent,” a GUI (graphical user interface) for non-engineers.

Local-AI Agent is a chat-based interface that automates a wide range of everyday business tasks through AI agents. When users provide instructions via chat, the AI agent autonomously handles everything from planning and execution to creating deliverables.

---

 Local-AI Agent

Key Functions

– Document search and analysis: Cross-document search and content analysis across folders
– OCR and document reading: Read PDFs, images, and documents
– Data processing and analysis: Process and aggregate structured and unstructured data
– Document generation: Create deliverables in Word, Excel, PowerPoint, and PDF formats
– Custom agent creation: Turn workflows into agents that can be shared and reused across the organization

Key Use Cases

●        Automated Drafting of Approval Requests and Reports (Financial Institutions)

By cross-referencing financial data in internal databases, past cases, and policy documents, the agent automatically generates draft approval documents based on the responsible staff member’s instructions. It autonomously performs tasks ranging from calculating financial indicators such as ROE and DSCR to industry comparisons, cash flow projections, and risk organization.

●        Digitization of Paper and Image Documents and Workflow Integration

Using OCR capabilities, the solution automatically converts analog materials such as paper forms, faxes, and scanned images into digital data, then processes them end-to-end by matching, aggregating, and analyzing them together with existing internal data. It automates data-conversion tasks that were traditionally manual, such as extracting parts lists from component drawings and aggregating costs in manufacturing.

---

IS Code

Solution Overview

IS Code is an on-premises coding agent for development engineers. Called from a terminal in the user’s code editor or similar environment, the agent autonomously executes each phase—from planning and specification design to implementation, editing, testing, and verification—based solely on chat-based instructions.
Its greatest advantage is that it enables agent-assisted coding without sending highly confidential internal code to the cloud. Compared with similar cloud-based services, processing speed can be 1.5 to 2 times faster while usage costs can be reduced to as little as one-fifth, based on measurements from our demonstrations.

Key Functions

– Code analysis and function summarization: Quickly understand existing code and accelerate specification review
– Implementation and editing: Generate code and edit existing code according to the plan, with diff display and snapshot saving support
– Testing and verification: Generate test cases, run static and dynamic tests, automatically create reports, identify bugs, and formulate remediation plans
– Skills function: Accumulate coding standards and naming conventions as reusable knowledge
– Memory function: Retain project information across sessions

Key Use Cases

●        Major Improvements in Development Productivity

Starting from reading design and specification documents, the agent supports the full development process, including code generation, test case creation, and proposing bug-fix strategies. Engineers can focus on core logic design and decision-making, significantly improving efficiency in repetitive implementation work.

●        Greater Efficiency in Code Review and Quality Assurance

By storing coding standards, security requirements, and review criteria as reusable knowledge, the agent can automatically provide comments and revision suggestions aligned with review perspectives. This helps prevent review quality from depending on individuals and raises the overall quality of the team’s output.

●        Modernization of Legacy Systems

The source code of core systems that have been running for many years is often poorly documented, and institutional knowledge is frequently lost when responsible personnel leave the organization. IS Code autonomously analyzes large volumes of legacy code, summarizes specifications, organizes dependencies, and proposes refactoring options.

---

Use and Validation of NVIDIA AI Infrastructure

Ippu Senkin is leveraging NVIDIA GPUs in its on-premises solutions and is working with NVIDIA on the application of GPU technology and information exchange. By optimizing state-of-the-art LLMs, including Nemotron, on the NVIDIA Blackwell architecture, the company has achieved inference accuracy comparable to frontier models and autonomous execution of multi-step tasks that goes beyond simple question-and-answer interactions.
In on-premises environments powered by NVIDIA GPUs, measurements from our demonstrations showed business processing speeds up to 1.5 to 2 times faster than similar cloud-based services.

Comment from Shuya Suzuki, Representative Director and CEO

By leveraging NVIDIAAI infrastructure, we can now provide state-of-the-art AI agent capabilities securely and at low cost without relying on the cloud. This solution significantly expands the potential for generative AI use in industries with strict security requirements, including financial institutions. We aim to create a world where dependence on specific individuals is reduced and organizational know-how can be accumulated and utilized as an AI agent platform.

About Ippu Senkin Co., Ltd.

Ippu Senkin Co., Ltd. provides AI application development and consulting services to help client companies improve operational efficiency and profitability. In particular, the company offers comprehensive support through services that include advanced business process assistance using generative AI, AI and data strategy formulation, and talent development, including in-house capability building.

Company Name: Ippu Senkin Co., Ltd.
Established: May 1, 2024
Representative: Shuya Suzuki, Representative Director, CEO/Founder
Address: 8-14-9 Ginza, Chuo-ku, Tokyo, Duplex Ginza Tower 8/14, Room 701
URL: https://ippu-senkin.com/