ML Research Engineer · MSc @ Tel Aviv University

Tom Ulanovski

I'm working at the intersection of machine learning and computational biology, with a focus on graph neural networks and large language models. I build pipelines that let deep learning research translate into real-world impact.

About

A bit about me

Tom Ulanovski headshot

I'm an MSc Computer Science student at Tel Aviv University, where my research sits at the intersection of graph neural networks, large language models, and computational biology. My thesis focuses on deep learning models for polyploid species evolutionary history.

Before TAU, I studied CS at Ben Gurion University (GPA 90.5, 3× Ashalim Program member) and spent most of 2024 as a Junior Data Scientist at RumenEra, building XGBoost models and preprocessing pipelines for multi-modal microbiome data.

I'm open for ML research engineer roles in industry, where research and engineering meet.

Outside of research, I spend a lot of my time with Panda, my dog, who keeps me busy enough outside the lab. When she lets me, I'm into basketball, snowboarding, surfing, and pretty much anything that gets the adrenaline going.

Research

What I'm working on

Improving LLM Embeddings via Inter-Layer Geometry

June 2025 – Present

ICLR 2026 GRAM Workshop (Tiny Paper Track) · First Author

A GNN-based method to enhance LLM embeddings for downstream NLP tasks by exploiting the geometry between transformer layers.

  • · Developed a GNN-based method to enhance LLM embeddings for downstream NLP tasks.
  • · Evaluated on 12 datasets across 9 LLMs, comparing multiple GNN architectures and baselines.

How Good Is Polyploid Phylogenetic Inference and Can We Do Better

November 2024 – Present

MSc Thesis · Tel Aviv University (in progress)

Large-scale benchmarking of polyploid phylogenetic inference methods, using 21 empirical datasets and extensive simulations as ground truth.

  • · Benchmarked 6 inference methods across 21 empirical datasets.
  • · Ran simulations under 9 evolutionary configurations per dataset, with the empirical datasets serving as ground truth for systematic method evaluation.
  • · Built an automated evaluation pipeline handling preprocessing, method execution, metric computation (edit distance, topology accuracy), and cross-method validation.

Experience

Education & work

Education

  • MSc Computer Science · Tel Aviv University

    • Thesis: How Good Is Polyploid Phylogenetic Inference and Can We Do Better.
    • Coursework: NLP, Deep Learning, ML with Graphs (GNNs), ML for Healthcare.

    November 2024 – Present

  • B.Sc Computer Science · Ben Gurion University

    • Thesis: Algorithms for Cluster Preservation Analysis in Adipocyte Subtypes.
    • GPA 90.5 / 100 · 3× Ashalim Program member (top ~5% of students).

    October 2021 – August 2024

Work

  • Junior Data Scientist · RumenEra , Beersheba

    • Built a Python preprocessing pipeline to normalize 15,000+ biological features (98% sparsity) using Rarefaction and CLR transformations, significantly improving model stability.
    • Developed XGBoost models to forecast farm productivity and methane emissions by combining animal health records with multi-modal microbiome DNA data.
    • Automated data pipelines merging multiple sources; used statistical analysis to quantify financial impact on farm operations.

    February 2024 – October 2024

Toolbox

Skills & languages

Languages

  • Python
  • Java
  • C++

ML / DL

  • PyTorch
  • PyTorch Geometric
  • HuggingFace Transformers
  • Optuna

Infra & Data

  • HPC cluster computing
  • Large-scale data processing
  • PostgreSQL
  • Statistical analysis

Languages

  • Hebrew · Native
  • English · Fluent
  • French · Fluent

Recognition

Awards & honors

  • 1st Place, BlockchainB7 Challenge
  • Rhodes Finalist, 2024
  • 3× Ashalim Program Member (top ~5%)
  • Edmond Safra Center for Bioinformatics Scholarship (2×)
BlockchainB7 Challenge prize

Contact

Let's talk

I'm open to ML research engineer and ML engineer opportunities, research collaborations, and interesting conversations. The fastest way to reach me is email.