Martín Gramática

Technical lead and AI lead in digital pathology. AI researcher and lecturer at UTN.

I build software that pathologists use every day, and I research how foundation models can find and segment cells in cytology slides.

Illustration of a cytology field: cells with dark nuclei, some outlined by segmentation masks
Nuclei in a Pap smear and the masks a model selects for them: the problem behind my current research on SAM2.

Work

Six years building digital pathology software at Smart In Media AG, where I now lead the technical side of the product line and the company's work with AI.

Smart In Media AGTechnical Lead & AI Lead, since 2021

Products

I own the architecture and roadmap of four applications used for diagnosis, remote collaboration and teaching, with a team of two developers, including real-time acquisition and processing of high-resolution microscopy images.

  • PathoZoom Scan & LiveView
  • LiveView Macro
  • RemoteScope
  • SmartGrossing

AI in the company

I am responsible for AI across the company: evaluating and integrating detection and classification models into our .NET applications, and building small in-house tools and proofs of concept.

Before: Computer Vision Engineer at Smart In Media (2020 to 2021), freelance computer vision projects, and seven years as a full-stack .NET developer.

Research

Since 2012 I have done research at the Software Research Laboratory (LIS) of UTN Córdoba on computer vision for digital pathology, mostly Pap smear cytology and histopathology. I also supervise student research teams.

Nucleus segmentation with SAM2

Learning which candidate masks to keep, evaluated on CCEDD. CONAIISI 2026.

Embeddings for assisted segmentation

What image embeddings reveal about cervical cytology. JAIIO 2026.

Foundation models without labels

UNI and BiomedCLIP to detect nuclei with no annotations. CONAIISI 2025.

  • 2026

    • Learned candidate mask selection for cell nucleus segmentation with SAM2: evaluation on CCEDD

      Gramática M.N., García M.A., Carrizo I. CONAIISI 2026, oral presentation.

    • Embedding analysis for assisted segmentation of cervical cytology

      García M.A., Gramática M.N.. SAIV, 55 JAIIO.

    • Relationship between object detection and image classification stages in Pap test

      Gramática M.N., Pianzola G., Alfici H.D., García M.A. Book chapter in Advances in Image Processing, Reliability, and Artificial Intelligence, Elsevier.

  • 2025

    • UNI y BiomedCLIP en la detección no supervisada de núcleos celulares en citología de Papanicolaou

      Gramática M.N., García M.A., Elmoni S., Ledesma L.O. XIII CONAIISI, Córdoba.

  • 2023

    • Intermediate task fine-tuning in cancer classification

      García M.A., Gramática M.N., Ricapito J.P. Journal of Computer Science & Technology 23(2).

    • Conv Pap Detect: enhancing conventional Pap smear analysis with deep learning for cell detection and classification

      Gramática M.N., García M.A., Ricapito J.P., Silva Fiezzi T. XI CONAIISI, Tucumán.

    • Clasificación de células en prueba de Papanicolaou en microscopía

      Gramática M.N., García M.A., Gignone M. CAIS, 52 JAIIO.

    • Intermediate task transfer learning and weakly supervised training in histopathology

      Gramática M.N., García M.A., Ricapito J.P. Conference on Cloud Computing, Big Data & Emerging Topics, La Plata.

  • 2022

    • Clasificación de tipos de cáncer en imágenes de histopatología

      García M.A., Gramática M.N., Rostagno L.A., Ricapito J.P. X CONAIISI, Concepción del Uruguay.

    • Clasificación de cáncer en histopatología: ¿texturas o detalles de alto nivel?

      Gramática M.N., García M.A., Rius B., Pucheta P., Luengo Ackermann R.B. JIS 2022, Buenos Aires.

    • Procesamiento automático de imágenes médicas

      García M.A., Gramática M.N.. XXIV WICC, Mendoza.

Teaching

I am a teaching assistant in the Artificial Intelligence course at UTN and design its integrative project. Every year students build a complete system around a technique industry is using right now, with free tools like Google Colab and Kaggle.

  1. 2024

    Object detection for chessboard analysis

  2. 2025

    Retrieval-augmented generation over text

  3. 2025

    Multimodal RAG over documents, figures and diagrams

  4. 2026

    Automatic scene description, from image to text

Earlier: the computer vision module of a data science diploma (2020 to 2021) and Mathematical Analysis II (2016 to 2018).

Education

MSc in Computer Vision, Robotics and Machine Learning

University of Surrey, UK, 2019. With Distinction, on a BEC.AR scholarship.

Information Systems Engineer

Universidad Tecnológica Nacional, Argentina, 2016.

Working on AI for pathology? Let's talk.

martingra@gmail.com

I am based in Villa Carlos Paz, Córdoba, Argentina, and work remotely with international teams. I speak Spanish and English, and some German.