We built these products for our own internal use. We also sell them to partners and clients with similar needs: as SaaS services, as white label together with other companies, or by integrating our solutions, or parts of them, into larger value propositions and technologies.
Want to know more about our products? Write to us
We start from a real need, ours or a client's, and take it to product in five steps, through to scale with capital and an industrial partner.
We start from market gaps, operational inefficiencies or emerging technologies (AI, automation, data analytics).
A lean MVP, tested internally or with beta customers, to validate product-market fit.
Proprietary stack, multidisciplinary team, agile method. AI-native products built to carry load.
We look for an industrial partner who has the problem we solved. We design the go-to-market with them, and they become the stakeholder of the launch.
We go to PE and VC only with a validated industrial partner in place. We connect financial capital to the industrial world.
Many start inside our daily operations and reach the market later.
If it works for us, it works for the market
We are the first users, so we constantly improve
We sell what we actually use
R&D investments amortized across internal use + external clients
We start from a problem AI makes solvable for the first time and carry it to a revenue stream. We pay for the research once, so every venture starts with proven technology inside it.
Literature, universities, reproducible experiments, public benchmarks. The lab produces reusable assets and writes down what worked, so no venture pays for the same lesson twice.
Research meets a real buyer. We take the problem inside organisations that have it, as consulting work, and check whether the technology holds, who buys it and at what price.
Technology that works and a clear buyer: we package it. Small team, agents and our own stack from day one, a business model picked from B2B licence, subscription, media and metered API.
The venture leaves as a company of its own. Deep tech when the value sits in the technology, tech service when it sits in the ability to deliver. KVA keeps a significant stake and stays in operations.
Inside, every venture climbs six rungs: ideation, validation, MVP test-to-learn, MVP test-to-confirm, scale 1, scale 2. Each rung carries the criteria that close it, and until they are closed you do not move up.
We built them in our own lab. What we learn on one product goes into the next.
Autonomous research and AI for science
Autonomous research with AI
A multi-agent system that starts from a claim, crosses disciplines, looks for evidence and counterexamples. A referee decides whether the result holds. CLI and web.
Users: Researchers and R&D teams
Fintech AI
AI suite for wealth management
Deal flow, wealth and private portfolio in one system. A conversational layer reads the documents, extracts the KPIs and proposes scenarios. Built with FNDX and FINDI.
Users: Family offices, advisors and private investors
Information intelligence
News, facts and information analysis
An epistemic framework that treats every news item as a multidimensional object: source, event, credibility, framing, propagation. It returns verifiable Reality Cards. Web, mobile and a B2B line.
Users: Newsrooms, fact-checkers and corporate communications
Audience intelligence
Synthetic interviews and product validation
Synthetic audiences to test products still in R&D: simulated interviews, value-proposition stress tests, workflows that turn thousands of signals into decisions. Focused on banking and insurance.
Users: Banks, insurers and product teams
Editorial generative AI
Agentic harness for editorial content
An agentic newsroom where humans keep the final say: signal radar, writing, fact-checking and publishing, then podcasts, reels and carousels from the same pieces. Text, audio and video in a single pipeline. KANSEI Magazine and the Lost in the Middle newsletter run on it.
Users: Newsrooms, newsletters and content teams
Ontologies and memory for agentic infrastructure
Deep tech memory for AI agents
Shared, governed memory of organisational decisions, exposed over MCP to the agents your team already uses. Deterministic kernel: it establishes which decision holds, with evidence and provenance, and abstains instead of guessing.
Users: Teams working with AI agents
Agentic infrastructure and organisational MCP
The operating brain of an organisation
An agentic terminal a company uses to build its own custom management software. Projects, tasks, documents and decisions live in one registry. The agents your team already uses read and update it over MCP. At KVA we run our projects, ventures and clients on it every day.
Users: Operations leaders and teams working with AI agents
Agents and AI tools for developers
Legacy code migration
A multi-agent system that reads whole codebases, rebuilds their dependencies and migrates COBOL, mainframe and .NET to modern stacks. 91.2% of tests passed on the first attempt on Poly-MigrationBench.
Users: Banks, insurers and large enterprises with legacy systems
Training, people development and AI literacy
Measure skills, train organisations
Measures skills by role, closes the gaps with a catalogue built around the company, verifies the effect with an AI tutor. In production with Gruppo Excellence, on a spin-off path.
Users: HR and L&D leadership at banks and large enterprises
There are three reasons: research gets a real proving ground, the team learns on systems in production, and the portfolio earns revenue independent of consulting.
Test emerging technologies in real environments (AI, automation, blockchain)
Develop entrepreneurial talent within the KVA team
Generate recurring value through diversified revenue (SaaS, services, licensing)
Demonstrate operational credibility to corporate clients and investors
Create scalable assets that can become standalone companies
Every venture leaves reusable assets: models, agents, pipelines. The next one starts from there.
Submit a co-creation project
Apply NowExplore the stack powering our ventures
View TechnologyLet's talk about building together
Contact Us
