Not one model for everyone.Millions of models,
one of them is Yours.
Intelligence must be shaped around your company, your processes and your taste, and placed entirely in your hands.
We imagine a future where the possibilities opened by technology and science become real capabilities, accessible and useful in the real world. Our mission is a single one: to build the competitive edge of those who choose to work with us.
What does competitive edge mean to us?
Giving those we work alongside the chance to stand out: to evolve their skills, their processes and the way they compete against rivals.
It takes a strategy, a long-term vision and a proprietary intelligence, tailored and improving over time.
Your data. Our method. Your model.
Hover over a step to follow its data.
What we build
Our research areas
A lab founded by master's students in Artificial Intelligence, with complementary skills. We study how to bring the latest models inside concrete problems: adapting them, testing them, making them reliable. We believe technology matters when it becomes a capability that is useful, verifiable and owned by those who use it.A lab of master's students in Artificial Intelligence: we bring the latest models into concrete problems, and make them reliable.
Master's in Data Science and Artificial Intelligence in Trieste, after a degree in Business Administration at Ca' Foscari. Passionate about startups and project manager with the VeniSIA accelerator in Venice.
Master's in Data Science and Artificial Intelligence in Trieste. Machine learning, full-stack development and AI applications. Passionate about science and frontier technology.
- 01Agents engineeringAgents that plan, use tools and finish long tasks: from researching the web to working at a computer like a person would.Agents that finish long tasks, like a person would.
- 02Post-trainingFine-tuning models and agents on your language, data and rules, so they answer the way your company would.Models that speak your company's language.
- 03Computer visionModels that read images and video: occupancy, flows, behaviour and anomalies, in real time.Images and video, read in real time.
- 04Unsupervised learningFinding structure where no one has labelled the data: clusters, anomalies and hidden risks.Structure and hidden risks in unlabelled data.
- 05Multimodal reasoningReasoning across text, images, maps and sensor data together, the way an expert does.Text, images, maps and sensors, together.
- 06Continual learningSystems that improve with use and stay reliable when the data around them changes.Systems that improve with use.
- 07EvaluationMeasuring what matters for your work, not for a leaderboard: before, during and after launch.Measuring what matters for your work.
- 08AI hardwareChoosing, installing and maintaining GPUs and infrastructure to run models on your own premises.GPUs and infrastructure on your premises.
A clear path, from the first conversation onwards.
Every step is designed to remove a reason to wait: you always know what happens next, what it costs and what you get.You always know what comes next, what it costs and what you get.
- Listen
We understand your context
We start from you: ideas, problems, expectations. Together we look at technical feasibility, the investment involved and the risks, before any commitment.Ideas, problems, expectations: we assess feasibility, cost and risk before any commitment.
- Research
Applied research
We compare architectures and models and make the trade-offs explicit: cost, performance, security, accuracy, speed. You see the alternatives and choose with full information.We compare models and trade-offs: you choose with full information.
- Proof
A demo before the contract
Before signing you see a working demo of what it will be, or you try a tool we have already built for a while. Then we agree on success metrics, timing and every step of the path.You see a working demo before signing, then we set metrics and timing.
- Build
Development, together
While we build, you stay involved: regular updates, tests and feedback. We correct course early, when it costs little. No black boxes.Regular updates and tests while we build. No black boxes.
- Deliver
Transparent delivery
We explain why every choice was made and how each part works. Security and data protection are designed in from the start, and everything comes with full documentation.Every choice explained, security from day one, full documentation.
- Train
Training your people
We train the people who will use it: what it is, how it works, how to use it and how it connects to everything else. So it becomes part of the organisation, not something foreign to it.We train the people who will use it, so it becomes part of the organisation.
- Care
Implementation and care
We put it into production and look after it: we monitor performance over time, as data and the organisation change, step in on faults and update models when better architectures appear.In production, monitored and updated over time.
- Stay
A partner over time
After release we remain your point of reference: for a question, a problem or a new project. We build relationships on trust and transparency.After release we remain your point of reference, for every new need.
4 elements, one system.
A project succeeds when they grow together.
Strategy
Knowing where artificial intelligence creates value for you: a few clear priorities, measurable goals and defined timelines. Fast projects, with a start and an end.
Built for your scale.
Every department has its own data and systems, under strict security and compliance requirements. We connect them into one system, on premises or in a private cloud, under your control.
- 1Agents and models integrated with your ERP, CRM and existing systems, department by department
- 2On-premises or private-cloud deployment, with access control, logs and audit
- 3Documentation for GDPR and the AI Act, monitoring and ongoing support
From the studio
DocumentintelligenceFrom documents to data, without retyping.OCR and language models to classify documents, extract fields and entities and validate them against rules and sources. The output is structured, and every value can be traced to the place in the document it came from.Designed for: finance and administration, professional firms, insurers, procurement, public administration.
Coffee andits alternativesIs there a market? Find out before building.For a VeniSIA food-tech startup weighing the launch of a legume-based alternative to coffee. Global, European and Italian markets compared, leading to entry barriers, positioning, launch scenarios and one clear answer: go ahead or not?Designed for: food-tech startups, food and beverage companies, venture builders and accelerators.
AutonomousAI agentsLong, repetitive work, handed to an agent.Agents that plan their steps, operate browsers, portals and business software with cursor and keyboard, and run research across multiple sources, citing them. They work in isolated environments, with checks at every step and human sign-off on sensitive actions.Designed for: back office and administration, customer service, market analysis and due diligence, portals and systems without APIs.
Your data, handled with care.
Those of us who build these systems every day see some things clearly. It is just as right that anyone entrusting us with their data should be cautious. So we would rather explain what we do, what can realistically happen and what measures to take, in plain words, before you ask.What we do, what can realistically happen and which measures to take, in plain words.
You decide where the data live. We explain the pros and cons of each option clearly.
The data sit at the centre, protected by several layers: if one gives way, the others hold.
- Your dataThey stay yours, inside your project.
- EncryptionData protected in transit and at rest.
- SandboxEvery agent in an isolated container, network closed unless needed.
- Minimal accessLeast privilege and role-based permissions; every access logged.
- ConfidentialityAn NDA signed before we see any data.
Before delivering we try to break what we built, and we keep checking it afterwards.
- 01Before release
Vulnerability scanning of code and dependencies (SAST, SCA) and configuration review.
- 02During testing
Red teaming: prompt injection, data leaks and unauthorised actions, fixed before production.
- 03After release
Continuous monitoring, central logs, alerts and an incident response plan.
- 01Isolated runtime
Every agent works in a sandbox with files, network and tools limited to what is needed, following the latest architectures for safe agents.
- 02Policies and guardrails
Explicit rules on what it can read, write and call; input and output filters against prompt injection and data leaks.
- 03Human checkpoints
Irreversible or sensitive actions (sending, payments, deletions) require human confirmation.
- 04Orchestrator and checker
An orchestrator agent splits the work, a checking agent verifies every result before it reaches you.
- 05External monitoring
A system separate from the agent watches its behaviour and stops it if it leaves its boundaries; permissions can be revoked in seconds.
- 06Traceability
Every action is logged and reproducible: who did what, when and why.
- Since 2018GDPR
Records of processing and impact assessments (DPIA) where needed.
- Aug 2026AI Act
In force: transparency, AI literacy and responsible use.
- Sep 2026Cyber Resilience Act
Reporting of actively exploited vulnerabilities and severe incidents.
- Dec 2027AI Act · high risk
Technical documentation, logs, human oversight and risk management.
Where it applies, NIS2 too. Every system comes with clear documentation: how it works, what data it uses, how it is monitored.
Some frequently asked questions.
It depends on the scope: after the first conversation you receive a clear proposal, with no obligation. Payment is split into phases, each tied to objectives agreed in advance, so you always know what you are paying for.
We compare several options on your real case and show you the trade-offs between cost, performance, security and speed. We have no preferred vendor: we recommend what works best for you.
You see a demo before signing, the success metrics are agreed in writing before we start, and you follow development step by step. Expectations and results are compared all along the way, not only at the end.
Your data stay yours, and so do the models trained on them. Where needed everything runs on your own infrastructure, and we sign a confidentiality agreement before seeing anything.
We treat personal data in line with the GDPR, document how every system works and prepare with you what the AI Act and the other applicable rules require in your case.
Yes: training is part of the project, not an extra. Every tool comes with documentation and a contact person who can answer questions.
We monitor performance over time and update the system when new models or architectures bring a real improvement. You stay at the state of the art without starting again.
No. Code, models and documentation are yours: you can continue with us, internally or with someone else. We prefer to be chosen again every time.