We build the data systems that AI runs on.
Main Abstraction is a data and AI consultancy in Lisbon. We build the pipelines that move your data, the search and retrieval systems on top of them, and the applications your team uses to work with both.
Lisbon · remote across the EU
GCP · BigQuery · AWS · Postgres · Python · SQL · TypeScript · Docker
Expertise
what we do
Data pipelines and cloud
Pipelines that were fine at launch get slower and more fragile as the data grows. We take over ingestion, scheduling, reprocessing and alerting, make them run predictably, and leave you able to see what they cost.
Retrieval and evaluation
Most teams can get an AI system to answer a question. Far fewer can say how often the answer is right. We build search and assistants over your own documents, and the evaluation that tells you whether they work before anyone relies on them.
On-chain systems
Contracts, and the data around them. We write and review Solidity, and build the indexing and analytics that make on-chain activity usable by the rest of your systems.
The Pipeline Review
A fixed way to start when you are not sure what to fix first. Twelve working days, fixed scope: we measure what your pipelines cost and how long they take, fix the biggest bottleneck, and hand it over.
Work
selected work
Applied AI · verification
A trust contract for generated reports
Every claim in a generated report carries a confidence label computed in code, from the reliability of its source. Not from the model rating its own output.
Read moreApplied AI · retrieval
Retrieval over a closed document collection
Question answering over a fixed corpus where every claim traces back to the passage it came from. Hybrid retrieval, reranking, and three verification checks that run in code after the model has finished.
Read moreBlockchain infrastructure
Access-control libraries for a zkVM execution layer
Two Rust libraries for a zkVM execution layer: an admin authority and a freeze authority. Security-relevant code, proven on-chain rather than argued for.
Read moreFAQ
questions
What do you work on?
Data pipelines and cloud infrastructure, and the search and AI systems built on top of them.
How does an engagement start?
With the Pipeline Review: twelve working days at a fixed price. We measure what your systems cost and how long they take, fix the worst bottleneck, and hand it over. Longer work continues from there.
What if I do not know what needs fixing?
That is normal, and the first three days answer it. Tell us what you are running and where it hurts.
What do I get at the end?
Working code in your repositories, running in your environment, and a written handover covering what changed and what we chose not to do.
Who decides the technical approach?
We propose, you approve. Anything affecting your platform, budget or security is agreed before it is built.
How do we work together day to day?
A weekly call and a shared channel. You can read the commits as they land.
How does pricing work?
Fixed price per engagement, agreed before we start. Day rates where a project needs them.
Is there a minimum commitment?
No retainer and no minimum term.
Who owns the code?
You do, on final payment. Cloud accounts and credentials stay in your name throughout.
Where does the code and data live?
In your cloud, under your accounts. No production data on our machines.
What happens after handover?
Two weeks of questions included. Monitoring and a runbook come with the work.
Where are you based?
Lisbon. We work remotely across Europe and travel for a kick-off when it helps.
How do we contract?
Portuguese company, EU VAT registered. Business to business inside the EU is reverse charge.
get in touch
Tell us what you are running, where it is slow or unreliable, and what a useful next step would look like. Twenty minutes on a call is usually enough to say whether we are the right people for it.
gon@mainabstraction.com