Commissioned Research & Development
We design and prototype technical ideas
before they become specifications.
Mindware Research Institute undertakes commissioned R&D for new AI products and core product capabilities. Our work is not general web production or routine chatbot development. We focus on problems that require an original combination of LLMs, machine learning, statistical analysis, databases, and domain-specific logic.
What It Is
Not consulting,
but the work of building testable hypotheses
We do not force a problem into an existing product. We examine its assumptions, design the structure it requires, and test its viability in a system that actually works.
Many AI projects begin by choosing the means in advance: use RAG, introduce agents, or assemble existing SaaS products. When building a new product or a core capability, however, the more important question is not which tool to select, but how the problem itself should be structured.
Mindware Research Institute brings together advanced-technology research, technical editing, self-organizing machine learning, statistical analysis, probabilistic models, databases, and LLMs to turn problems that have not yet been standardized into working research prototypes.
Research Scope
Areas of Research and Development
We do not sell a fixed package for a particular industry. We design the core mechanism around the client’s specific technical problem.
Knowledge systems beyond conventional RAG
We design knowledge systems that combine conceptual structures, routing, multiple knowledge sources, and decision rules rather than merely retrieving documents and generating answers.
Concept formation from text and data
Using embeddings, self-organizing learning, clustering, and network structures, we model patterns that are difficult to capture with fixed taxonomies.
Long-term and episodic memory for AI
We prototype memory mechanisms that structure experience and make it available for later judgment and recommendations, rather than merely storing conversation history.
Integration of qualitative and quantitative data
We integrate text, open-ended responses, categories, and numerical data into a common analytical structure to identify new segments and decision-relevant patterns in operational data.
Decision support and hypothesis diagnosis
We design decision-support functions that combine evaluation criteria, probabilistic models, rules, and data analysis instead of relying on LLM-generated text alone.
Original systems that standard tools cannot produce
We integrate LLMs, machine learning, statistics, databases, and external APIs to research and implement original capabilities at the core of a product.
Project Fit
Projects that fit—and those that do not
We accept only one or two projects per year and prioritize work with genuine originality and room for technical investigation.
Suitable Projects
- You need to determine the technical architecture of a new AI product or core capability
- Existing RAG or chatbot approaches do not provide sufficient quality
- You need to integrate text, numerical, categorical, or other heterogeneous data
- You have a research idea but no working prototype
- You need an architecture that keeps internal data from leaving your environment
- The problem is not yet defined well enough to hand a specification to a conventional development company
Projects Outside Our Scope
- Standard corporate websites or e-commerce sites
- Projects whose only goal is to install an off-the-shelf chatbot
- Staff augmentation, time-and-materials staffing, or on-site development
- Routine business-system development with fully defined specifications
- Projects centered on implementing a large number of screens in a short period
- General AI adoption consulting or briefings on fashionable technologies
Development Process
A phased path from research to product prototype
Rather than committing to a large development contract at the outset, we review the results of each phase before proceeding.
Technical Problem Review
We review the background, users, data, existing systems, and constraints to identify the central problem that should be treated as R&D.
Concept Study
We investigate relevant technologies and alternative approaches, then design the data structures, processing flow, evaluation method, and scope of the MVP.
Working MVP
We implement the central hypothesis with limited data and functionality, then test accuracy, speed, usability, and technical limits.
Product Prototype
Building on the MVP, we add the user interface, APIs, authentication, data management, evaluation functions, and operational controls required for a product prototype.
Evaluation, Handover, and Next-Step Decision
We organize the validation results, remaining issues, and operating conditions, then determine the next path: internal development, handover to a conventional development company, or continued joint development.
Deliverables
Deliverables that support the next decision—not just a proposal
Each phase produces a technical conclusion and a working deliverable. Even when an implementation does not succeed, we identify the obstacle and the conditions under which the approach could become viable.
Working Principles
Principles of Commissioned R&D
Developed as a new client-specific system
Commissioned projects are not simply deployments of ConceptMiner or ThinkNavi. We design a new architecture and codebase around the client’s specific problem.
Intellectual property and usage rights defined by contract
We distinguish client-specific deliverables, pre-existing proprietary technology, and reusable development know-how, and define ownership, usage rights, and permitted reuse in the contract.
Designed around confidentiality and data governance
At an early stage, we confirm NDA requirements, data location, whether external APIs may be used, and whether local or on-premises processing is required.
Frequently Asked Questions
Frequently Asked Questions
Can we discuss a project before the specifications are defined?
Yes. In fact, the Concept Study is intended for situations where a technical question exists but has not yet been translated into a specification.
How much does a project cost, and how long does it take?
Pricing is estimated individually for each problem and phase. A typical Concept Study takes three to four weeks, an MVP one to two months, and a project extending through product prototyping approximately three to six months in total.
Do you work with companies outside Japan?
Yes. We accept projects that can be researched, designed, reviewed, and developed remotely. Contract terms, language, time-zone differences, and data-transfer conditions are confirmed in advance.
Can you provide maintenance and operation after development?
Continued development after the research prototype is considered individually. When large-scale operation and maintenance become the primary requirement, we generally plan for handover to an appropriate development company.
Must a project use ConceptMiner or ThinkNavi?
No. Those products demonstrate our research capabilities, but commissioned projects are designed as new systems around each client’s specific problem.
What information is needed for the initial discussion?
Please describe the problem to be solved, the intended users, available data, limitations of the current method, and any constraints that must be respected. A completed specification is not required.
Begin the discussion before the specification exists.
We will review the technical problem and its context to determine whether it is suitable for commissioned R&D.