Skip to content
Case studies · In production

Real systems, in production.

Proven capabilities, explained from the inside. Client details are left out for confidentiality; how each system works is described in full.

01
01
Calculation engine

Expert knowledge as an API

The rules that used to live in spreadsheets, now an auditable service.

Problem

Business-critical knowledge trapped in legacy spreadsheets: hard to audit, impossible to version and dependent on whoever maintains them.

Solution

An API with validated, versioned rules, traceability for every run and a catalog that grows without touching the core. Any system in the company can query it.

What you see today

  • Every result is traced: what was calculated, with which data and which version
  • The catalog grows without touching what is already validated
  • Where it fits: technical calculations, rate setting, quotes, scoring

Origin: electrical engineering · standards-based calculation in production

The challenge

IEC/UNE standards living in spreadsheets: impossible to audit, version or reuse across projects.

Architecture
APIServerlessInfrastructure as codeAI engine
Result

Every run is traced, and the calculation catalog grows without touching the core.

02
02
Document generation

Documents that write themselves

From your data to the final document, in seconds.

Problem

Repetitive documents written by hand—reports, proposals, technical memos—with copy errors and endless review rounds between the person who knows and the person who writes.

Solution

An engine that takes the data, validates it and returns the final document ready for review: always the same format, stable formulas, zero manual transcription.

What you see today

  • Same data, same document: zero transcription errors
  • Exhaustive validation against reference templates
  • Where it fits: proposals, recurring reports, contracts, technical documentation

Origin: engineering technical memos · in production

The challenge

Technical memos written by hand in Word, with formulas that broke with every revision.

Architecture
JSON requestValidationCalculation engine.docx render (OMML)
Result

A Word-ready document in seconds, with no intermediate PDF pipeline.

03
03
Market monitoring

Automated market monitoring

Opportunities arrive on their own, ranked.

Problem

A team manually tracking dozens of sources—portals, bulletins, official websites—with false positives, coverage gaps and summaries that arrived too late.

Solution

A monitor that queries the sources daily, scores each opportunity against your criteria, uses AI to summarize only what matters and delivers it where the team already looks: email, Telegram or Slack.

What you see today

  • Coverage: official portals, bulletins and multilateral bodies, extendable by country
  • Quality filter: only the highest-scoring opportunities consume AI resources
  • Where it fits: tenders, competitors, pricing, grants

Origin: international tenders · daily delivery in production

The challenge

Manual tracking of dozens of public sources, with false positives and summaries that arrived too late.

Architecture
Serverless ingestionHeuristic scoringLLM summaryEmail + Telegram
Result

Only relevant opportunities come through, ranked: the filter acts before the model is called.

04
04
Private AI

Private AI for your team

Built into where your team already works, with the data in your own environment.

Problem

The team needs conversational AI, but with real control: deployment in their own cloud, integration with the internal stack and data terms that off-the-shelf products did not offer for their case.

Solution

Assistants with memory, versioned capabilities and a standard connection to their tools. Deployed on the client's own infrastructure: the data stays in their account.

What you see today

  • It remembers the context: sessions and persistent memory
  • Capabilities tested end to end, with a quality threshold before they are switched on
  • Where it fits: internal support, operations, company knowledge

Origin: a technical assistant in Slack for engineering · the client's cloud

The challenge

Conversational AI without giving up data privacy or forcing changes on the internal stack.

Architecture
Persistent sessionsVersionable skillsStandard connectorClient tenant
Result

Assistants that are measured and tuned against real use, with the data always in your environment.

05
05
Web design

This very website

The storefront, treated as a product.

Problem

Corporate websites are usually templates with no engineering behind them: slow, unmeasured, inaccessible and with forms that nobody knows whether they deliver.

Solution

In-house design on a modern web framework: audited performance and accessibility, technical SEO, cookie-free analytics, a form with verified delivery and continuous deployment to the cloud.

What you see today

  • You are reading it right now: every improvement ships the same day
  • Audited exhaustively: design, copy and performance
  • Where it fits: your company's website, built to the same standard

Origin: sertoria.com · the case study you are browsing

The challenge

Making the storefront live up to what it sells: if the website promises engineering, the website itself is the first proof.

Architecture
Custom designModern web frameworkContinuous deploymentCookie-free measurement
Result

A fast, accessible, measurable website — built with the very service it sells.

Your case?

The next one could be yours.

If your problem looks like any of these, the first conversation comes with no commitment.

Let's talk