Nicolas Cozzarin

Nicolas Cozzarin

AI Product Owner for regulated industries

I have 10 years of experience in tech, from backend development and QA to product management. Today I work on AI and data products in finance and the public sector, where data protection and compliance are part of the job.

  • Based in France, open to relocation (Switzerland, France, Spain) or remote
  • EU citizen, can work in Switzerland without sponsorship
  • English, French, Spanish, German

About

I'm a Product Owner with 10 years of experience in backend development, QA, functional analysis and product management. In the last few years I have focused on AI and data products.

In Geneva I worked at Union of Financial Corners, a money transfer and currency exchange company, where I was product owner for three products, among them crypto purchase in the branches and an online exchange platform. For my degree internship I built an NLP system for the office that investigates police misconduct in Córdoba, Argentina. It removes personal data from each complaint and suggests a category, and a prosecutor makes the final decision.

My current focus is AI compliance, AI ethics and alignment. I want to help organisations use AI and automation where it really makes their work better, and not only companies: associations, NGOs and public institutions often have fewer resources to do it and a lot to gain. I have started doing this as a volunteer, for example by designing the AI framework for the Hack the Fork hackathon.

Alongside my professional work, I have volunteered extensively in animal welfare and animal rights, and I am now exploring the potential impact of artificial intelligence on the situation of animals, a crucial topic that remains largely undiscussed. This includes examining the present and potential use of AI in factory farming and, conversely, in reducing wild animal suffering, as well as how reducing speciesist attitudes in frontier AI systems could shape outcomes for animals now and in the future. Finally, I consider how AI tools could help boost our efforts in animal advocacy.

Experience

  1. IT Product & Project Coordinator (Product Owner role)

    Oct 2022 – Oct 2025

    Union of Financial Corners SA, Geneva, Switzerland

    Money transfer and currency exchange company with branches in six Swiss cities, working under FINMA and anti-money-laundering rules.

    • Product owner for three products: crypto purchase in the branches, an online currency exchange platform, and an improved Western Union money transfer flow for branch clients.
    • Coordinated a team of about 10 developers and external suppliers. I wrote the user stories and acceptance criteria and prioritised the backlog.
    • Ran quality and compliance tests on new products, services and processes before they went live, checking them against anti-money-laundering and operational requirements.
    • Coordinated suppliers, subcontractors and communication agencies for the launches, and managed SEO and social media.
  2. AI Product Owner & ML Developer

    Nov 2025 – Jun 2026

    Security Forces Disciplinary Control System, Government of Córdoba, Argentina

    Degree internship, remote and part-time, with the Coordination of Information Registry and Analysis.

    • Owned the NLP system from start to finish: gathered the requirements with the legal staff, wrote the specification, defined the roadmap and built the model.
    • Designed the anonymisation pipeline (regex, spaCy NER and Microsoft Presidio) so that no personal data reaches training or inference, in line with GDPR and data protection law.
    • Fine-tuned BETO (Spanish BERT) with a class-weighted loss after testing TF-IDF baselines. On a blind test set of 2,189 complaints it reached 77% recall on institutional violence and an F1 score of 0.72 on gender-based violence.
    • Designed the review step with role-based access and an audit log: prosecutors see the confidence for each category and approve or override every suggestion, and their corrections are used to retrain the model.
  3. Functional Analyst

    Feb 2019 – Dec 2022

    FlyDevs, USA (remote)

    • Translated business requirements into functional specifications and user stories for the development teams.
    • Worked as the link between stakeholders, UI/UX, QA and engineering during the whole product delivery.
    • Facilitated the Agile ceremonies and helped unblock the team when delivery got stuck.
    • Helped streamline the release process by aligning development and CI/CD practices with business priorities.
  4. Lead QA Analyst

    Mar 2017 – Dec 2019

    Charly Inc., Canada (remote)

    • Led quality assurance on several software projects running at the same time.
    • Defined the test strategies and coordinated a distributed QA team to make sure functional and technical requirements were met.
    • Set up the processes for functional, integration and regression testing.
    • Kept the test documentation and defect tracking up to date, and improved how we detected defects before release.
  5. Python Backend Developer (Odoo)

    Feb 2015 – Jun 2016

    E-MIPS, Buenos Aires, Argentina

    • Developed and customised Odoo modules in Python to automate business processes such as CRM, inventory and accounting.
    • Integrated Odoo with external applications and APIs, including ERP systems and payment gateways.
    • Documented the solutions and used Git and Docker throughout development.
    • Worked in Agile delivery cycles to maintain and improve the applications.

Education

  1. BSc in Artificial Intelligence and Robotics

    2021 – 2026

    Universidad Siglo XXI, Argentina

    Grade 8.40/10, diploma expected in December 2026. Specialisation in artificial intelligence, data science and robotics, with coursework in machine learning, deep learning and generative AI. Final degree project: a validator that checks EU AI Act controls in the deployment pipeline (see Selected work).

  2. Exchange in AI and Robotics (Master's-level courses)

    Nov 2024 – Nov 2025

    Hochschule Furtwangen University, Germany

    Courses in machine learning, deep generative models, computer vision, robotics and microcontroller systems, with a lot of hands-on work in Jupyter. I also passed the German B2 exam.

  3. Electronics Technician Diploma

    2009 – 2015

    Instituto Técnico Industrial San Judas Tadeo, Argentina

    A base in electronics, applied systems and hardware, and a very project-based way of solving problems.

Selected work

EU AI Act checks in the deployment pipeline

Final degree project, Universidad Siglo XXI, 2026

A validator built to run as a step of a GitLab CI pipeline. For each new version of a credit model, it measures whether the model approves men and women at different rates and checks five controls of the EU AI Act. If one fails, the deployment is blocked and the report gives the rule and the article.

Results: Working prototype. In a first run, version 1 of a test credit model passed the five rules and version 2 was blocked by the fairness rule (demographic parity difference of 0.15, limit 0.10). The evaluation against manual review takes place in November 2026.

Stack: Python, RDF, SHACL, GraphDB, Fairlearn, GitLab CI, Streamlit, Docker

The consultancy and the bank are a modelled case, not real clients.

Crypto sales in Swiss branches

Union of Financial Corners, Geneva

Clients pay cash at the counter and receive Bitcoin or Ethereum in their own wallet, or sell crypto for cash. I led the launch: the requirements, the choice of the Swiss crypto partner, the backlog with the external developers, the testing, the branch procedures and the staff training.

Results: Live in all branches in Geneva, Lausanne, Bern and Zurich. Clients come back regularly, we had every document the auditors asked for, and the service is being extended to partner kiosks across Switzerland.

Anonymised complaint triage for police oversight

Government of Córdoba, Argentina, 2025–2026

An NLP pipeline that removes names, ID numbers and addresses from citizen complaints before anyone reads them, and then suggests one of the office's eight official categories. A prosecutor approves or changes each suggestion.

Results: On a blind test set of 2,189 complaints, 77% recall on institutional violence and an F1 score of 0.72 on gender-based violence. Everything runs on the office's own hardware.

Stack: Python, PyTorch, BETO (Spanish BERT), spaCy, Microsoft Presidio

The code is not public because the system works with government data.

Justice Inequality Simulator

Deep Generative Models course, Hochschule Furtwangen, 2025

A way to test if a model that predicts court outcomes is biased. I change one sensitive detail in the case description, like gender or ethnicity, and measure how much the prediction moves.

Results: In several cases, swapping a sensitive word moved the predicted probability by 0.10 to 0.25. Test accuracy of 0.78, probably optimistic: the write-up explains the limits of the evaluation.

Stack: Python, Legal-BERT, PyTorch, scikit-learn

All posts

Research interests

These are the topics I want to keep working on, at work and in a master's.

AI and animals
How AI is used today in factory farming, how it could help reduce wild animal suffering, and how speciesist attitudes in frontier AI systems could affect animals now and in the future. I'm also interested in how AI tools can help animal advocacy. It is an important topic that is still barely discussed.
AI in public institutions
I want to understand how courts, regulators and oversight bodies can use AI without it becoming a new black box. In Córdoba that meant a person reviewing every decision, an audit log and keeping the data on government hardware. The EU AI Act asks for many of the same things for high-risk systems.
Bias in AI models
In the Justice Inequality Simulator I tested bias by changing one detail in a case and checking if the prediction changed. I like this kind of test because anyone can run it on a system they don't control, including regulators and journalists.

Leadership and volunteering

  1. AI Enabler, Hack the Fork

    2026

    Paris

    I designed the AI framework for the hackathon and helped test the AWS environment. During the event I mentor the teams on how to scope their AI projects.

  2. Organiser, We The Free

    2026 – present

    Angoulême

    I organise the local chapter of an animal rights group. I manage the materials budget, organise outreach and social events, and recruit new activists in the area.

  3. Logistics, Vegan Campout Argentina

    2021 – 2022

    Buenos Aires

    I coordinated stall logistics and suppliers for an event with 2,000 attendees, helped with the programme, and managed a team of 20 volunteers on the day.

  4. DevOps Volunteer, Difusión V and Voicot

    2019 – 2022

    Remote

    I set up and maintained the hosting and CI/CD for the organisation's three websites.

  5. Organiser, Buenos Aires Animal Save

    2019 – 2022

    Buenos Aires

    I co-organised vigils outside slaughterhouses. I opened each event with a briefing for the attendees, was the contact with the slaughterhouse workers, supported participants during the vigils and produced footage for social media.

  6. Organiser, Anonymous for the Voiceless

    2018 – 2020

    Buenos Aires

    I created and ran Cube of Truth street events with 20 to 60 activists. I coordinated the teams and materials, and welcomed and briefed new members.

Skills and languages

Product
Agile (Scrum, Kanban), backlog management, user stories, requirements, roadmapping, stakeholder management, Jira, Confluence
AI and data
Python, PyTorch, Hugging Face Transformers, scikit-learn, pandas, spaCy, NLP (classification, NER), SQL, LLM APIs (OpenAI, Anthropic), RAG, AI agents, local LLMs
Governance
Data protection by design, anonymisation and pseudonymisation, GDPR, EU AI Act (human oversight), AML and FINMA compliance, audit logging
Engineering
Docker, CI/CD, Git, REST APIs, React, TypeScript
Languages
Spanish (native), English (C1), French (C1), German (B2), Chinese (HSK 3)

Courses and certifications

  • Fundamentals of Agents

    Hugging Face Agents Course, Sep 2026

    Unit 1 of the Hugging Face Agents Course: what AI agents are, how large language models work as their brain, how agents use tools to act on their environment, and the think, act and observe cycle.

    Show credential

  • Future of AI

    BlueDot Impact, Aug 2026

    An introduction to what AI can do today, where it is going over the next decade, and how to start contributing to a better future.

    Show credential

  • Introductory workshops on effective altruism

    Altruisme Efficace France, Jun–Jul 2026

    High-impact strategies and evidence-based approaches to global problems.

    Show credential

  • Mendix Rapid Developer Certification

    Mendix, Apr 2025, certificate number 90924

    Certification in fast, scalable low-code application development.

    Show credential

  • MLA Training Module (anti-money laundering)

    Union of Financial Corners, Oct 2022

    Training on anti-money-laundering regulations and financial compliance.

    Show credential

  • Build ReactJS Applications

    Udemy

    Hands-on development of dynamic, component-based front-end web applications.

  • Web Development

    Coderhouse, Jun 2020

    Fundamentals of modern web architecture, responsive design and deployment.

  • Backlink Management Course with Greg Gifford

    Semrush

    SEO, link-building and optimising an online presence.

  • Advanced C++ Programming and Modern Practices

    Programming course

    Memory management, object-oriented design and high-performance code.

  • Hardware Description Language for VHDL Development

    2017

    Programming and logic design for complex digital circuits and FPGA integration.

  • University Course in Microelectronics

    2016

    Design, fabrication principles and practical uses of microelectronic systems.

Contact

The best way to reach me is by email. References are available on request.