Marcel Claramunt — CV

AI & Full-Stack Engineer [LinkedIn] [GitHub] [Website] [Email]

Skills & Technologies

  • Languages: Python, TypeScript
  • Machine Learning: PyTorch, TensorFlow, OpenCV (with a focus on computer vision)
  • Backend: FastAPI, Postgres, CosmosDB, Azure Blob Storage
  • Frontend: React, Vite, NextJS
  • DevOps: Docker, Kubernetes, Azure
  • Sales & Customer Support: cold outreach and customer success

Professional Experience

Solo Founder, Moveread (2023 — Present)
  • Built an AI-powered scoresheet digitation for chess tournaments
  • Sold over 15 tournaments across Spain, France, Germany and Brazil, generating 50-150€ of revenue per tournament
  • Collected a dataset of >500k annotated handwritten chess moves and >5k annotated grid images (for table structure recognition)
  • Trained a Table Transformer to >95% accuracy, allowing fully automated preprocessing for most scoresheet models
  • Trained an OCR model to >90% accuracy on handwritten chess moves
  • Finetuned a postprocessing algorithm to >99.5% precision and 50% recall (allowing human-level, completely automated processing of half the games)
Independent Open-Source Projects (2023 — Present)
  • Haskellian: a functional programming library [repo, docs]
  • Pipeteer: a simple and explicit durable execution framework [repo, docs]
  • Cit: a developer productivity bash CLI tool, providing a centralized way to scaffold projects, open repositories, run services, and more [repo, docs]
  • KV: an async key-value interface for Python, with implementations for SQL, Redis, Azure Blob, Cosmos DB, the file system, and more [repo, docs]
Plus many more open-source contributions and projects on my GitHub profile

Education

BS in Computer Science
  • Faculty of Informatics, Polytechnical University of Catalonia (UPC)