rrRoman RosetData · Systems · Product
enesca
Barcelona20+ years of engineeringSenior engineer · Accent Obert

I turn complex problems into useful systems.

I'm Roman Roset, a senior hands-on engineer with more than twenty years of craft. I connect a real need with the architecture, data and an operable product: from backend and process through to applied AI.

→ I bring technical judgement, delivery capability and one useful obsession: technology should make people more capable, not more dependent.

See projectsView résumé
scopefrom problem to production
specialismdata · systems · AI
practicejudgement · code · operations
Roman Roset
Track record
20+ years · product, data & research
Professional profile

The value I bring

A senior, hands-on profile: I can understand the problem, choose the architecture, build the system and support the team until the outcome actually works.

01systems judgement

I see the whole without losing the detail

I model the domain, data, services, risks and operation before fixing a solution. Then I go down to the code and test whether the architecture survives the real world.

Track record · backend · data · research · product · operations
02delivery

I take a need all the way to product

I translate non-technical goals into contracts, flows and deployable software. I can lead the decision and remain hands-on with the code.

Examples · from clinical research to products and live services
03reliability

I decide with evidence

Contracts, tests, metrics, cost and privacy are part of the design. With AI I compare models and results; I do not confuse a convincing answer with a reliable system.

Method · baselines · E2E · real cost · human validation
04hands-on leadership

I grow teams and judgement

I provide structure, boundaries and context so people can make better decisions. I prefer lightweight processes that leave a trace and reduce personal dependencies.

Experience · tech lead · co-founder · process · 23 university projects
A rule I set myself

“I am good at giving structure to complex problems; I also know I can build more than they need. That is why I work with short milestones, real evidence and one rule: complexity has to earn its place.”

Way of working

How I approach projects

I combine technical depth, product judgement and AI tools without turning the fashionable tool into the architecture.

01 / understand

Problem before solution

I clarify the domain, users, invariants and what “works” will mean. I cut before building and separate what is essential from what is merely interesting.

domainproductriskprivacy
02 / build

Systems that can be operated

Backend, data, contracts, tests and deployment are one decision. I build systems to remain understandable, observable and changeable after the first release.

PythonFastAPITypeScriptPostgreSQLDocker
03 / amplify

Applied AI with judgement

I build lightweight harnesses: bounded context, reusable skills, checks and the right model for the task, quality and cost. I use agents only when coordinating contexts or specialisms creates more value than complexity.

harnessskillsmodel selectionMCPevaluation
Selection

Where I create value

My current work is private. I describe the kind of problem and responsibility, not internal details. Personal projects and research complete the picture.

Current work · confidentialinternal systems

From structured data to useful decisions

I have moved from building databases, ETL and BI to designing ways for teams to query, interpret and turn that data into decisions with AI support. My responsibility spans the data foundation, answer quality, cost and fit with the human process.

structured dataMage · Metabaseapplied AIAPIsevaluation
foundation
Data · quality · context
supports
structures
system
Query · report · interaction
assists
enables
people
Judgement · decision · action
decide
Current work · confidentialintegration and internal product

Business process without tool friction

I design tailored integrations, automations and interfaces over CRM and ERP systems. The goal is not another software layer, but adapting HubSpot, Odoo or Zoho to the actual process and reducing the learning curve for the people using it.

HubSpotOdooZohoAPIstailored frontends
Continuous learning

Selected credentials

The track record demonstrates the level; these credentials show that I keep tools and foundations current. I completed nine Anthropic Academy programmes in 2026; these are the most relevant.

Spring Certified Professional

Java · Spring · backend

VMware2023
Verify

AI Fluency: Framework & Foundations

Score 9/10

Anthropic Academy2026
Verify

Building with the Claude API

Model and product integration

Anthropic Academy2026
Verify

Introduction to Model Context Protocol

MCP · tools and context

Anthropic Academy2026
Verify

AI Fluency for Nonprofits

Applied AI for organisations

Anthropic Academy2026
Verify
Career

Twenty years of craft

From supercomputing and bioinformatics to digital health and data-driven innovation. Always with the same compass: technology useful for people.

2023 — present

Accent Obert

Senior engineer · systems, data & applied AI
End-to-end design and implementation of internal decision systems over structured data, CRM/ERP integrations and AI solutions shaped by quality, cost and real utility.
2020 — 2023

Fundació .cat

Data engineer, backend & BI
Data-driven innovation, backend of the Business Intelligence platform and orchestration of cross-platform integrations.
2017 — 2019

WellTech TMS

Co-founder & Software Engineer
Therapeutic software spin-off. Technical leadership adapting academic research into a commercial product.
2011 — 2019

Universitat Rovira i Virgili · ALGOS

eHealth lead & Software Engineer
Technology strategy and apps for chronic pain patients: Fibroline and Painometer.
2011 — 2015

Universitat Oberta de Catalunya

Adjunct Professor · Computer Engineering
Undergraduate and master teaching and supervision of 23 final projects.
2002 — 2009

BSC · INB · Arte · MicroArt

Research, semantic web & bioinformatics
Supercomputing on MareNostrum, clinical ontologies and agent-based decision support systems.
About me

A simple idea drives me: technology should make people more capable, not more dependent.

I don't aim to revolutionise industries, but to add my grain of sand so people achieve the results they want —and feel more competent doing it. I believe in constant learning, active listening and collaboration.

With AI, I look for the smallest structure that produces a reliable outcome: a skill, a model or a harness; agents only when the problem genuinely requires coordination.

Català
Native
Español
Native
English
Intermediate · B1
"Sharing is learning."

This site is open on GitHub, like most of what I build. I like making things others can reuse, question and improve —the same logic I apply to teams.

— Roman Roset Mayals