Silvio Martinico headshot

Silvio Martinico

Research Engineer @ Weaviate | PhD Researcher @ UniPi & ISTI-CNR
Email Scholar ORCID GitHub LinkedIn X

I'm a computer scientist with a Math background. I work on making search faster and more scalable, one of the real bottlenecks in AI right now.

I'm wrapping up a PhD focused on efficiency in Information Retrieval and Vector Search, and at the same time working as a Research Engineer at Weaviate, because I wanted my work tested against problems that only show up at production scale.

During my PhD I co-founded tusKANNy, a research group built around one goal: making vector search efficient, and shipping the fastest open source and production-ready retrieval libraries.

Open Source Projects

  1. TACHIOM
    A state-of-the-art multivector retrieval system based on Token-Aware Clustering (TAC), hierarchical indexing, and late-interaction-optimized Product Quantization, written in Rust with Python bindings. It retrieves over 600M ColBERTv2 embeddings (MS MARCO) in 10 ms on a single core. Built on top of kANNolo and Vectorium, and integrated into PyLate.
    GitHub
  2. kANNolo
    A flexible Rust library combining state-of-the-art indexing techniques for both dense and sparse embeddings. Designed to make prototyping new ANN algorithms fast and ergonomic. Delivers 2x+ faster retrieval than FAISS at equal accuracy and is integrated into PyTerrier.
    GitHub
  3. Vectorium
    A Rust library for storing, accessing, and compressing dense, sparse, and multi-vector embedding datasets. Provides a unified dataset/encoder interface shared across tusKANNy's indexing and search crates. Includes an exhaustive search API and a CLI tool for ground-truth computation.
    GitHub
  4. Awesome Multivector Retrieval
    The reference collection of papers, code, models, and datasets for the late-interaction multivector retrieval community, covering ColBERT-style models, indexing techniques, software libraries, and benchmarks. Featured in the PyTerrier tutorial at SIGIR 2026.
    GitHub

Publications

  1. Lossless Compression of Proximity Graph Indexes for Efficient Approximate Nearest Neighbors Search
    Silvio Martinico, Francesco Benocci, Leonard Dahlmann, Francesco Gargiulo, Tobias Heuer, Franco Maria Nardini, Cosimo Rulli, and Rossano Venturini
    ACM International Conference on Information and Knowledge Management (CIKM), 2026
  2. Efficient Multivector Retrieval with Token-Aware Clustering and Hierarchical Indexing
    Silvio Martinico, Franco Maria Nardini, Cosimo Rulli, and Rossano Venturini
    ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 2026
    PDF DOI Code
  3. Multivector Reranking in the Era of Strong First-Stage Retrievers
    Silvio Martinico, Franco Maria Nardini, Cosimo Rulli, and Rossano Venturini
    European Conference on Information Retrieval (ECIR), 2026
    PDF DOI Code
  4. Efficient Approximate Nearest Neighbor Search on a Raspberry Pi
    Silvio Martinico, Franco Maria Nardini, Cosimo Rulli, and Rossano Venturini
    ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 2025
    PDF DOI Code
  5. kANNolo: Sweet and Smooth Approximate k-Nearest Neighbors Search
    Leonardo Delfino, Domenico Erriquez, Silvio Martinico, Franco Maria Nardini, Cosimo Rulli, and Rossano Venturini
    European Conference on Information Retrieval (ECIR), 2025
    PDF DOI Code

Experience

Education

  • PhD in Computer Science @ University of Pisa
    Pisa, Italy · November 2023 – April 2027
    Research on efficient algorithms and data structures for Information Retrieval and Approximate Nearest Neighbor (ANN) search, carried out at the ISTI-CNR HPC Lab. Focus on late-interaction / multi-vector retrieval, compression, and index design. Published at SIGIR, ECIR, and CIKM.
  • MSc in Computer Science @ University of Pisa
    Pisa, Italy · October 6, 2023
    Graduated 110/110 with honours. Thesis on Graph-based Indexing for Efficient and Effective Approximate Nearest Neighbors Search.
  • BSc in Mathematics @ University of Pisa
    Pisa, Italy · May 14, 2021
    Thesis on the decidability of the word problem for the free exponential ring and field.
    Thesis

Work

  • Research Engineer @ Weaviate
    Remote · August 2026 – Present
    Developing and optimizing multivector retrieval systems (ColBERT-style) as part of the Applied Research team.
  • Founding Member & Core Maintainer @ tusKANNy
    Pisa, Italy · April 2024 – Present
    Lead developer of the group's open-source retrieval libraries. I take algorithms from paper to production, covering Rust implementation, benchmarking, API design, documentation, and releases.
  • Research Associate @ ISTI-CNR
    Pisa, Italy · November 2023 – Present
  • Visiting Researcher @ University of Tokyo
    Tokyo, Japan · 02/2025 - 03/2025
  • Teaching Assistant @ University of Pisa
    Pisa, Italy · 03/2024 - 06/2024
    Tutor for the Algorithms and Data Structures course within the M.Sc. program in Data Science.

Out of Office

Beyond research and systems engineering, you can find me lifting at the gym, casually playing beach volleyball or basketball, traveling, or grabbing a Spritz or a beer with friends. I'm a massive One Piece fan, I survive mostly on pasta, and I write my fastest Rust code while listening to heavy metal and Italian rap.

Contact