PL EN

Independent data and AI engineering

Every pair, considered.

We build data platforms and AI integrations side by side with your team, and leave systems it can run itself.

Book a call

More data than answers.

The data sits in the CRM, the warehouse, a few spreadsheets and a folder of PDFs. The reports disagree, and the AI pilot ended as a slide.

Tools are rarely what's missing. What's missing is checked relationships: between systems, between numbers, between the board's question and the data that answers it.

That's where we start.

A hand with a marker sketches a data architecture on a whiteboard: boxes joined by arrows, several crossed out.
Fig. 1. The architecture drawn on a whiteboard, before the first line of code.

Four variables make six pairs.

Our mark is half of a correlation matrix. Each square is one pair of variables.

  1. We test every pair before we trust the data.
  2. Most pairs are noise.
  3. One pair changes the decision. We find it, then build the system around it.

Fig. 2. Illustration on sample data. Variables: delivery time, discount, support tickets, customer churn.

What we build.

Six practices and one aim: data the board trusts, and AI that works in everyday operations.

  1. AI in your processes

    Document classification and extraction, fuzzy matching, AI assistants connected to your CRM and data warehouse, MCP servers for internal software.

    Claude · Model Context Protocol · AWS · Azure

  2. Data platforms and warehouses

    Warehouse design for sales, CRM and risk. Migrations to cloud data warehouses, high availability and Snowflake best practice.

    Snowflake · SQL Server · PostgreSQL · Teradata · SQL Server Always On

  3. Integration and pipelines

    ETL and ELT pipelines you can test, version and rebuild from scratch.

    dbt · Airflow · Fivetran · Matillion · Informatica · AWS Glue · Azure Data Factory

  4. Data quality and monitoring

    Quality rules and checks, data-flow monitoring, alerts, and automatic insights written by AI.

    Amazon CloudWatch · Amazon SNS · alerts on Slack and email

  5. Cloud and infrastructure as code

    Serverless architecture on AWS and Azure, defined in Terraform and shipped through CI/CD. Infrastructure as code in every deployment.

    AWS · Azure · Terraform · GitHub Actions · Azure DevOps

  6. Reporting and advisory

    Reports that arrive on Slack and email by themselves. Leading data teams, architecture reviews and training.

    Power BI · Tableau · SSRS

Selected work.

Two engagements, described without client names. Details and references are available on request.

International private equity firm

Assets under management (AuM)
over USD 400 bn
Assets under administration (AuA)
over USD 275 bn

PDFs that route themselves into the right systems.

An automated pipeline takes in PDF documents and delivers their data to the systems that need it. AI works at every stage, including our own fuzzy matching, which compares what a document says with what the system already holds.

  1. 1

    Classify

    An AI model recognises the type of each PDF.

  2. 2

    Extract

    Data pulled out against the schema for that document type.

  3. 3

    Verify

    Checked against the data already in the system, with our own fuzzy matching.

  4. 4

    Distribute

    Updated data sent on to several downstream systems.

Fig. 3. Serverless architecture on AWS, defined in Terraform and deployed through GitHub Actions, monitored in CloudWatch with notifications via SNS.
AWS S3 · AWS Lambda · AWS Systems Manager · Amazon SNS · AWS Glue · Amazon CloudWatch · Terraform · GitHub Actions

US legal services firm

Annual revenue
over USD 100 m
A desk with a stack of printed documents marked up in pencil, reading glasses and a laptop.
Fig. 4. Documents and notes from a data architecture review.

A data architecture that adds up, and AI for the whole organisation.

We started by reverse-engineering the existing data architecture. Then came data quality and flow monitoring with automatic AI insights, Claude connected to the CRM, the Snowflake warehouse and the monitoring tools, custom MCP servers for internal software, and training for the team.

Reports now arrive on Slack and email by themselves.

Claude · Model Context Protocol · Azure · Azure Data Factory · Snowflake · Terraform · GitHub

Projects in 15 industries

  • Private equity
  • Legal services
  • Banking
  • Pharmaceuticals
  • Life sciences
  • Manufacturing and mining
  • Energy and fuels
  • Automotive
  • Construction
  • Electronics
  • Telecoms and media
  • Cloud and IT
  • Transport and telematics
  • Recruitment and HR
  • Fitness tech

How we work.

Every engagement is run by a pair: one of our consultants and someone from your team. When we finish, the know-how stays with you.

Two people at one desk, seen from behind: one points a pen at a chart on a laptop, the other takes notes.
Fig. 5. Working as a pair at one desk.

  1. Diagnose

    We map the systems, the data flows and what they cost, and say what to fix first and why.

  2. First release

    One use case taken all the way to production, not to a slide.

  3. Build

    Platform, pipelines, AI integrations and monitoring. All of it as code, in your repository and your cloud.

  4. Hand over

    Documentation and training, until your team runs the system on its own.

Let's talk about your data.

Tell us in a few sentences what you're dealing with. We'll reply with a time for a short call.

Book a call
or write directly to [email protected]

Company
Jakub Suszczyński Join It
Tax ID (NIP)
5223145732
Address
ul. Morcinka 4 lok. 10
01-496 Warszawa, Poland
Email
[email protected]

© 2026 pairwisedata. This site uses no cookies and no analytics.

Wersja polska