AI software and training

Software for your data. Training for your people.

Software: automations, data pipelines and models, handed over working on your data. Training: online courses, workshops and team sessions. Fixed price and the must-haves agreed before work starts.

Book a free 30-minute call

Software builds in 2 to 6 weeks from kick-off
Fixed price, fixed time
You own what we build, on full payment

Free, 30 minutes, no preparation, in Portuguese or English. Within 2 business days of the call you get a one-page pitch to agree, or a straight “not a fit”.

Prefer to write? Use the form or email hello@xval.ai

Software

Advice on where AI fits and tests that measure it before you commit; then the build itself, handed over working on your data.

One clear workflow goes straight to a pitch and a fixed-price build. Several candidate tasks or a stalled AI project start with a Discovery Sprint; an open question about accuracy or integration starts with a test. No specific task yet, only a mandate to use AI? Start with training for your team.

Automate

AI automations

Work that arrives as documents, emails, photos and tickets, handled by AI inside your systems. For example: purchase orders typed into the ERP, supplier files checked against the required list, field photos screened for anomalies, product descriptions written from ERP data. It can also answer plain-language questions from your records. It connects through an import file, or through your system’s API or database where one is available; we confirm which before pricing. A person checks only the ambiguous cases.

Software you own instead of a subscription

If your team uses a small, stable part of an expensive tool, we rebuild that part around how you work. It follows your data model and your workflow, uses AI where it helps, and connects to your systems. Delivered in weeks for a fixed price; after migration you retire or reduce the subscription. Not a copy of the product: the part of the job your team actually does. On full payment you receive the rights to what we build; third-party software and models stay under their own terms.

Data and models

Data pipelines and ETL

Getting data out of the ERP, CRM, spreadsheets and files into one place you can rely on. Pipelines run on schedule, databases are enriched from external sources, and deduplication and validation make the numbers agree with each other.

Dashboards and metrics

The metrics the business runs on, defined once and kept current. Dashboards your team opens every morning, analysis of what moved and why, and answers to specific questions from your own data.

Machine learning models

Models trained on your history: forecast demand, score leads, flag risk, decide which cases to look at first. Built against a baseline and an evaluation set, delivered running inside your systems, answering every week without anyone asking.

Advise and test

Discovery Sprint

1 week · credited against a build started within 90 days

For several candidate tasks, or when we need your data in hand to see how the work is done. We sit in on the next real run of the task, such as this week’s review or this month’s batch, and do it with you, partly automated on your own exports. You keep that result and every rule written down, the main ones in code. You also get what data you can export yourselves, a baseline from recent history where it exists, and a pitch for the next step. Useful even if you stop there.

Test before you commit

1 day to 2 weeks · go/no-go in writing

Before we promise a number, we measure it on your data. A Feasibility Sprint (2 weeks for each AI step) builds an evaluation set: real cases from your documents with the correct answer. It tries the approaches against a pass bar, the minimum result agreed in writing, and reports the result, the failures and the cost per case. An Integration check (1 day) tests whether we can read from or write to your system, with a written pass or fail. If it is a go, the measured number becomes the build’s acceptance test: the result it must reach before it is done.

Training

Courses, workshops and team sessions, designed and delivered: we produce the course your learners take, or teach the people in the room.

Three formats

Courses for your learners

One pilot module, or a full course of six

An online course produced from sources and learning outcomes agreed at kickoff, delivered as files ready to upload to your course platform. Each module has short narrated video lectures, worked examples and a 10-question quiz with feedback. You review each module once; we do not host or administer your platform. Start with one module to test the format with your learners before you commit to a full course. On full payment the course content is yours; the avatar and voice services used for narration stay under their own terms.

Workshops

Half a day to 2 days · up to 12 people

Sessions for your learners, members or partners, built around exercises on real cases and, on request, your own data. Run as a stand-alone session or alongside a course, for the people who want to apply what it teaches.

Training for your team

Half a day to 2 days · up to 12 people

For the people who will use AI at work: what it can and cannot do, and how to use it safely with company data. They also learn to spot the next task worth automating. On request, the examples are built from your own documents.

Courses, workshops or a session for your team: Book a free 30-minute call

How it works

One call, a one-page pitch you agree, then a build, course or workshop at a fixed price. Every software build includes 3 months of Care; after that, a monthly plan or a full handover.

01
Agree
Free call · pitch within 2 business days
On the call you describe the task and the data, or who needs to learn what; we say whether it is a fit. Within 2 business days you get a one-page pitch: the problem, the must-haves we will deliver, and what is left out. We agree it with you on a short call before any price. If shaping needs your data in our hands, or an unknown has to be measured first, the pitch starts with a Discovery Sprint or a test. Nothing starts before you sign.
02
Build
Fixed price · must-haves agreed in writing
Software: we build the must-haves agreed in the pitch, on your data. Smaller builds run on demand from files; larger ones are deployed where they will run, preferably in your own cloud account. Each build ends when it passes the acceptance test agreed before work starts, with a handover pack: code in your repository, runbook, evaluation set, credentials list and a one-hour session. Courses: audience, outcomes and sources agreed at kickoff, then produced module by module, each module reviewed once by you, delivered ready to upload. Workshops and team sessions: prepared for your audience and run on the agreed dates.
03
Operate
Software · 3 months of Care included, then monthly
Every software build includes 3 months of Care, our maintenance plan: hosting administration, monitoring, security updates and small fixes, so you see it working before you choose. Then pick a monthly plan, cancellable with 30 days’ notice, or run it yourselves or with another supplier using the handover pack. Hosting and AI model usage are always paid separately.

Who you work with

Pedro Marcelino, founder

PhD in machine learning; research in forecasting and decision analysis; publications cited around 780 times on Google Scholar.
Leads every xval.ai engagement personally and answers for every deliverable, from discovery and training to data, automation and software.
The person you meet on the call is the person accountable for the result. No sales hand-off.

Verify: LinkedIn · Google Scholar

Find out if your project is a fit.

In a free 30-minute call we look at what you need: the task and the data behind it, or who needs to learn what. Within 2 business days you get a one-page pitch to agree, or a straight “not a fit”. If a question is still open, the pitch starts with a Discovery Sprint or a test.