AI & Systems Engineering
Enrico Peruffo
Software architecture, infrastructure, local AI, orchestration and technical implementation. Turns complex hypotheses into executable and verifiable systems.
Enrico Peruffo × Michele Valleri · Bergamo
We design systems around real work.
We step into processes, understand where time and information are lost, then build the digital system that is actually needed — software, automation and AI only when they add real value.
00 / Who we are
Meet the team
Product & Process Design
Michele Valleri
Process analysis, product design and the transformation of operational problems into systems that can actually be built.
01 / Projects
Projects delivered
Every project starts from an existing process. The outcome is not a technology stack: it is a different way of working.
Job and installation operating system
From information split across folders, office, workshop and site surveys to one system that follows each job from measurement to installation.
Jobs · Field → workshop · Shared data · Automation
Operations platform for a psychology practice
One operating system for patients, professionals, appointments, sessions and administration, while preserving distinct roles and responsibilities.
Operations · Scheduling · Roles · Administration
Operations platform for a pet facility
Customer booking and staff operations in the same system: stays, availability, tasks, grooming, documents and payments.
Hospitality ops · Booking · Customer area · Daily operations
Request qualification and quoting system
From unstructured requests that were hard to evaluate to a guided flow that collects the information needed before preparing a quote.
Lead flow · Guided request · Website · Human review
02 / AI Lab
A lab for turning AI hypotheses into verifiable systems.
Here we develop and test AI projects that do not start from a client brief: infrastructure, products and research directions built to find out whether an idea can become a real system.
System Twin
Reconstructing real digital systems into an evidence-backed semantic model spanning code, configuration, APIs, runtime, telemetry and documentation.
Question: Can we reconstruct what a system does without conflating observed facts, declarations, derivations and inferences?
Research / model-definition · schema and stack not frozen
Local AI
A public core for describing, validating, executing and evaluating bounded local-AI work.
Question: How do we prove what a local model can do before authorizing it to perform real work?
Public Core 0.3.1 RC1 · schema 1.0.1
Opportunity Engine
An evidence-first orchestrator that starts from observable operational problems and turns them into controlled research, qualification and commercial opportunities.
Question: Does starting from problems instead of company lists produce prospects that human reviewers consider materially better?
Current milestone: Problem Research → Prospect Research → Qualification
Trustworthy Reasoning
Feasibility research into cognitive AI that keeps evidence, time, versions and contradictions explicit inside the reasoning process.
Question: Can we build a defensible scientific project around reasoning, abstraction and planning that preserves the history of evidence?
Phase 0 — Call decomposition and research governance
Evolving AI prototypes, experiments and projects · we make the direction public, not the details that form the technical advantage.
03 / FAQ
Questions worth clarifying.
The point is not to apply more technology. It is to understand what change is actually needed, what is worth building and which boundaries should remain explicit.
What kind of problems do you work on?
Situations where work depends on manual handoffs, scattered information, disconnected tools or decisions that require people to reconstruct context over and over. Before proposing a solution, we identify where the operational friction actually starts.
Do you always start with AI?
No. AI is an option, not the starting point. If traditional software, integrations or deterministic automation solve the problem better, we prefer the simpler system. We use AI when it materially changes what the product can do.
Do you build everything from scratch?
Almost never by principle. We first assess what already works, preserve mature specialist tools and build the missing layer around them. The value is not rewriting everything; it is making the overall system work better.
Can you work with data that needs to stay inside the company?
Yes. When privacy, data ownership, latency or control justify it, we design local or on-premise architectures and workflows where data does not need to leave the organization. Cloud remains a tool, not a requirement.
How does a project start?
We reconstruct the real process: people, handoffs, tools, data, constraints and decisions. Then we identify the smallest change capable of removing a meaningful friction, make it verifiable and only then expand the system.
Can AI Lab projects become startups?
Some can. The Lab exists to separate an interesting idea from a technical and product thesis that actually holds. When a direction demonstrates enough value, it can evolve into a standalone product, spin-off or new venture.
Are you open to partnerships or investment?
For selected directions, yes — especially when the counterpart brings capital, distribution, data, domain expertise or scientific and industrial validation. The site shows the thesis and the problem; details that form part of the operational advantage remain private.