# Federico Marcos

![Federico Marcos](/resources/cognition/portraits/federico-marcos.webp)

Data Infra & Analytics Lead @ Cognition | Previously Data Scientist & Engineer @Scale AI  
Cognition  
Montevideo, Montevideo, Uruguay

## At Cognition

- Role: Data Infra & Analytics Lead (Engineering)
- Joined: Mar 2024
- How they joined: Directly
- Before Cognition: Senior Software Engineer, Scale AI
- Overlapped before Cognition with other Cognition people at: Scale AI (https://www.autumn.ai/resources/cognition/worked-together)

## Identity
- LinkedIn: https://www.linkedin.com/in/federicomarcos
- X: @marcosfederic
- GitHub: https://github.com/marcosfede
- Domain: cognition.ai

## Notes
- X post, verbatim: "Proud to say I built the DeepWiki MCP. Standing on the shoulders of giants!" DeepWiki's official MCP server (announced by Silas Alberti in May 2025, featured by OpenAI) is his work. (https://x.com/marcosfederic/status/2021662430951293374)
- Cognition describes DANA (Data Analyst Agent) as a specialized Devin that "connects to our data warehouse through MCP" and answers Slack questions like "why did signups drop Tuesday?"; given his MCP and analytics-blog work and Data Infra & Analytics Lead title, DANA sits squarely in his area. (https://cognition.com/blog/how-cognition-uses-devin-to-build-devin)
- LinkedIn post (2025-07-23), verbatim: "Happy to ship what I've been building for the last couple of weeks, Devin now connects to all of your favorite tools!" and "Feeling very bullish on MCP!" He built Devin's MCP (Model Context Protocol) integration/Marketplace, which launched the same day. (https://www.linkedin.com/posts/federicomarcos_happy-to-ship-what-ive-been-building-for-activity-7353889276440276992-oyP1)
- He reposted Scott Wu's Jan 2026 note on 'Dana' (Devin ANAlyst) as a team-wide win for running usage and analytics queries. (https://x.com/marcosfederic/status/2010167221054795974)
- X post, verbatim: "Today I shipped something different from code: a blog post on turning Devin into your personal data analyst!" alongside a LinkedIn post: "Wrote a really cool blogpost showing how to get Devin to answer all your analytics questions today, pulling insights from all your data sources right from Slack!" (https://x.com/marcosfederic/status/1961131274023965049)
- Cognition colleague Theodor Marcu credited the Devin MCP launch to "Sara Xiang + Federico Marcos + Jeffrey Ling really cooked with this launch", and tied it to the Windsurf deal: "And with Windsurf, we'll cook even harder than ever before". Sara Xiang and Jeffrey Ling are his closest visible launch collaborators. (https://www.linkedin.com/posts/theodormarcu_sara-xiang-federico-marcos-jeffrey-ling-activity-7353898027507441664-wGZO)
- His most-starred open-source repo is marcosfede/algorithms (1130 stars): "Solved algorithms and data structures problems in many languages", the visible trace of his algorithmic/competitive-programming practice. (https://github.com/marcosfede/algorithms)
- He is listed as a contributor to CognitionAI/metabase-mcp-server, Cognition's open-source Metabase MCP server. (https://github.com/CognitionAI/metabase-mcp-server)
- He joined Cognition in Mar 2024, the month Devin was publicly launched, and was there through the July 2025 Windsurf acquisition. A colleague's launch post the same week said "with Windsurf, we'll cook even harder"; nothing public shows him moving to Windsurf work. (https://www.linkedin.com/posts/theodormarcu_sara-xiang-federico-marcos-jeffrey-ling-activity-7353898027507441664-wGZO)
- At Scale AI he wrote mongoose-query-logger, a "Mongoose plugin for query execution logging and explaining"; his LinkedIn says he created an open-source tool to log and monitor expensive and slow MongoDB queries while tech-leading Scale's customer dashboard. (https://github.com/marcosfede/mongoose-query-logger)
- Public headline says "Data Infra & Analytics Lead @ Cognition | Previously Data Scientist & Engineer @Scale AI". (https://www.linkedin.com/in/federicomarcos)
- X post about Cognition's culture, verbatim: "When someone asks what the bar is like at Cog, I'm gonna link to this tweet". (https://x.com/marcosfederic/status/2095592178160480754)
- GitHub (marcosfede, joined 2015, 58 public repos) bio: "Data Wizard @ Cognition / Passionate about Physics, Math, Music and Technology", company @usacognition, location Uruguay. (https://github.com/marcosfede)
- Swiss army knife data wizard obsessed with supporting important decisions with actual data and deep diving into ML models. Everything can be made intuitive with the right abstractions and visualizations.

I've led million-dollar projects, single-handedly owned production critical services end to end, optimized acquisition channels, experimented with A/B tests, and always enjoy building new production-ready models. 

I have a strong math and analytics background and a passion for algorithms and ; this comes from the profile summary and preserves the person's own positioning. (https://www.linkedin.com/in/federicomarcos)
- On X he follows Cognition leaders and peers (Scott Wu, Steven Hao, Silas Alberti, Jeffrey Ling, Russell Kaplan, Varun Mohan of Windsurf) and agent-tooling builders. That follow list points to the internal network he works with, including the Windsurf founder. (https://x.com/marcosfederic)
- Co-presented 'Federico Marcos, Bruno Cascio. Celerative.' at UNLP's VII Jornadas de Cloud Computing & Big Data (2019). (https://jcc.info.unlp.edu.ar/en/vii-jornadas-de-cloud-computing-big-data/)
- Pre-Scale he worked on cualbondi.com.ar, a free-software public-transport routing platform (4k+ DAU; Django/PostGIS/Vue/Go), as full stack engineer and DevOps 2018-2020. (https://www.linkedin.com/in/federicomarcos)
- Contributed to celerative/PPS-NN-Gabriel with GabyMorgi, an intern research project (Jupyter); matches his LinkedIn claim of supervising 10+ ML interns at Alto/Celerative in La Plata. (https://github.com/celerative/PPS-NN-Gabriel)
- Previous roles include Founding Software & Data Engineer at Cognition; Senior Data Scientist at Scale AI; Senior Software Engineer at Scale AI; DevOps Engineer at Alto; Technical Lead at Alto; Machine Learning Engineer at Alto; Full Stack Developer at Celerative; Full Stack Software Engineer at cualbondi. (https://www.linkedin.com/in/federicomarcos)
- Education history includes Bachelor of Science (BS), Physics, Universidad Nacional de La Plata; Certificate in Full Stack Web Development, Computer Software Engineering, freeCodeCamp. (https://www.linkedin.com/in/federicomarcos)
- Location is Montevideo, Montevideo, Uruguay. (https://www.linkedin.com/in/federicomarcos)

## Experience
- Data Infra & Analytics Lead at Cognition (Jan 2026 - Present)
- Founding Software & Data Engineer at Cognition (Mar 2024 - Jan 2026)
 #snowflake #redash #tableau #dbt #dagster #SQL #plotly

- Constructed multiple multivariate classification and regression ML models as part of my daily DS workflow. Studied correlations of user-based features to worker performance, long-term kpis, bad quality / spam predictors, etc. Analyzed syntactic and semantic diversity metrics for Generative AI’s top customers with NLP
#scikit-learn #pandas #xgboost #jupyter #seaborn #stanza #NLP #spacy

- Owned the company’s migration to Google Analytics 4. Ingested and created ETL pipelines for user-based real-time events. Root caused multiple acquisition channel misconfigurations and bottlenecks. I integrated new channels like Facebook, Reddit and Bing Ads. I highlighted campaign costs and performance for every campaign in a master dashboard I created. 
#segment #google-ads #google-analytics #snowflake #redash #etl

- Designed and carried out A/B experiments to see the impact of business-critical decisions based on data from real users.  
#pandas #numpy #seaborn #jupyter #stats #python
- Senior Software Engineer at Scale AI (Apr 2020 - Feb 2022): - Tech lead for the main customer-facing dashboard for scale AI. Managed to improve performance by 5-10x while shipping dozens of new features and fixing outstanding technical debt. Optimized our backend endpoints and created an open-source tool to log and monitor expensive and slow Mongodb queries

- Co-designed the redesign of one of the most critical services for scale infrastructure: the task queue. The new system powers today’s scale’s labeling services and handles millions of tasks each month, with improved performance, monitoring, debuggability, tooling, and custom per-project logic. The codebase transformed from a complexity nightmare into a robust, well though-out, intuitive, and well-documented system

- Member of the initial SWAT-like team that developed the initial POC for scale retail and e-commerce solutions. This was at the time, scale’s most important deal ever. We turned this incredibly ambitious MVP into life in less than three weeks and gained the customer’s trust, growing the team over time to hundreds of people and closing multiple million-dollar deals for the company. I was promoted for my contribution to this project.

- Single-handedly owned and executed multiple projects end to end. From initial specing, drafting the system design, syncing with stakeholders, carrying out the implementation, monitoring, dashboarding, and presenting results to management.

#typescript #mongodb #javascript #nodejs #puppeteer #aws #react #datadog #snowflake
- DevOps Engineer at Alto (Dec 2019 - Apr 2020): - Architecture design for an online betting platform.

- Developed a fault-resistant in-premise infrastructure.

- Used tools like Jenkins, Docker, Docker Swarm, Ansible.
- Technical Lead at Alto - La Plata, Argentina (Mar 2018 - Apr 2020): - Manager of the development team.

- Progress tracking and division of tasks according to scope.

- Handling Requirements and specification of objectives for the different projects.

- Owned architecture and software design decisions.

- Ensure quality in the development of products and in the service provided.
- Machine Learning Engineer at Alto (Oct 2017 - Apr 2020): - Team Leader in diverse ML and Data science projects for international clients

- Built a Recommender System for a fitness app with more than 4M users (200k+ MAU). 

- Gave several conferences in the ML area representing my company.

- Project Manager and Supervisor for 10+ interns, researching a wide range of Machine Learning topics including image recognition, classification using deep Convolutional Neural Networks, Genetic algorithms and neural networks for game AI, Natural language processing, Sentiment Analysis, Recommender Systems.
- Full Stack Developer at Celerative - La Plata (Feb 2017 - Mar 2018): - Project architecture design, AWS devops, pixel-perfect frontend development, development of REST/GraphQl microservices

- Developed multiple end to end applications interacting with clients directly.
- Full Stack Software Engineer at cualbondi (Apr 2018 - Apr 2020): Full Stack Software Engineer at https://cualbondi.com.ar, a free Software platform for public transport routing.

4k+ Daily active users, expanding to several other countries

Some of the tech involved: Python / Django / asyncio, Node.js, Postgresql / Postgis, Vue.js, golang, PWA

Responsabilities: Fullstack developer, Devops, server maintenance, CI/CD
- Full Stack Developer at BlueDraft - Soluciones Analíticas - La Plata, Argentina (Sep 2017 - Sep 2018): Full Stack Developer. Team Lead on both frontend and backend for the BlueChip application, a BI platform for managers and sales teams that allows to speed up the analysis, forecasting and reporting processes of economic indicators

Developed a SPA built using Javascript, React and Redux along with a backend REST API using Django and Python
- Full Stack Developer at Savana, Inc. (Feb 2017 - Sep 2017): Full Stack Developer. Complete UI remake of the main application used by the US based fintech client. 

Automated the process of converting screens to the new design using python scripting and we were able to reach the ambitious objective in time

Some tech used: Python, Javascript, node.js, electron.js, web scrapping, bootstrap CSS
- Math Tutor in English at SMARTHINKING, Inc. (Sep 2014 - Feb 2017)

## Education
- Universidad Nacional de La Plata - Bachelor of Science (BS), Physics - (Jan 2007 - Jan 2016)
- freeCodeCamp - Certificate in Full Stack Web Development, Computer Software Engineering - (Jan 2016 - Jan 2016): Full Stack developer

## Sources
- https://www.linkedin.com/in/federicomarcos
- https://www.linkedin.com/posts/federicomarcos_happy-to-ship-what-ive-been-building-for-activity-7353889276440276992-oyP1
- https://x.com/marcosfederic
- https://cognition.com/blog/ai-data-analyst
- https://cognition.com/blog/how-cognition-uses-devin-to-build-devin
- https://www.linkedin.com/posts/theodormarcu_sara-xiang-federico-marcos-jeffrey-ling-activity-7353898027507441664-wGZO
- https://jcc.info.unlp.edu.ar/en/vii-jornadas-de-cloud-computing-big-data/
- https://x.com/thenateherman
- https://www.linkedin.com/posts/federicomarcos_data-analytics-with-coding-agents-activity-7366895511049506816-bZGx
- https://api.github.com/users/marcosfede/repos?per_page=100&sort=updated
- https://api.github.com/users/marcosfede
- https://github.com/CognitionAI/metabase-mcp-server
- https://x.com/marcosfederic/status/2021662430951293374
- https://x.com/marcosfederic/status/2010167221054795974
- https://x.com/marcosfederic/status/1961131274023965049
- https://github.com/marcosfede/algorithms
- https://github.com/marcosfede/mongoose-query-logger
- https://x.com/marcosfederic/status/2095592178160480754
- https://github.com/marcosfede
- https://github.com/marcosfede?tab=repositories
- https://github.com/celerative/PPS-NN-Gabriel


Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-11-04. Profile: https://www.autumn.ai/resources/cognition/people/federico-marcos · Everyone at Cognition: https://www.autumn.ai/resources/cognition/people