Kia ora G’day Hello

Finding structure
in complex data.

I’m Ilia. I build computer vision systems at inDrive and independent research software — connecting applied mathematics, causal dynamics, and scientific computing.

Senior Data Scientist @ inDriveData scientist · Researcher · Builder
Field notes / 001Unlabelled sheep
Counting sheep: a sketch of structure in dataA synthetic point cloud drawn as a flock of sheep. Reveal flocks to colour them using a small k-means demonstration.

K-means sketch · synthetic sheep

At the intersection of
Computer visionApplied mathematicsScientific computing

01 / Selected work

Big questions. Working code.

Independent projects at the boundary
of research and engineering.

SpaceDataLab

Research initiative

Precision, from orbit to observation.

Astrodynamics in Rust, with the verification as the deliverable: orbit propagation and determination, GNSS, constellation design, and Earth-observation geometry. Grounded in the primary literature.

  • Rust
  • Astrodynamics
  • Earth observation

CausalDyn

Research software

Learn the dynamics. Make better decisions.

Causal learning meets physics-structured models and constrained control. Research tools and benchmarks that evaluate both prediction quality and the decisions that follow.

  • Causal inference
  • JAX
  • Optimal control

Tools for the working scientist

Rust performance. Python ergonomics.
01 / Clustering

betula-cluster

Turn large embedding datasets into compact, numerically stable microclusters. A Rust core with a scikit-learn-style Python API.

  • Rust + Python
  • Streaming
  • Machine learning
02 / Indexing

lexindex

Compact string-to-ID indexes built in Rust. Fast lookups, prefix and fuzzy search, with memory-mapped storage and Python bindings.

  • Rust + Python
  • FST / MPH
  • Data infrastructure

Further explorations

All repositories

02 / A bit about me

Curiosity, with
an implementation.

I’m Ilia Gradina, a Senior Data Scientist specialising in Computer Vision at inDrive. My background brings together applied mathematics, software engineering, and machine learning.

Since joining inDrive in 2021, I’ve built and developed computer vision systems across document understanding, identity verification, vehicle recognition, and visual content moderation.

Documents & OCR
Multilingual text recognition, document parsing, and vision-language models.
Faces & liveness
Face detection and recognition, image quality assessment, and liveness detection.
Vehicles & number plates
Detection and recognition for cars, motorcycles, and commercial vehicles, alongside number-plate recognition.
Visual moderation
Detecting inappropriate content and advertising in images.

I work across the full ML lifecycle: dataset design and annotation, model training and evaluation, reusable libraries, inference optimisation, and production deployment.

I started in space ballistics and satellite systems software, then moved to data science at Kantar: NLP, marketing mix modelling, and voice and video analytics. That mathematical and engineering foundation carries into my independent work today: causal dynamics, astrodynamics, and scientific tools in Python and Rust.

I’ve also contributed to Fedora as a package maintainer since 2014. My COPR repositories provide builds of Maxima and FriCAS — computer algebra tools that connect naturally with my interest in mathematical software.

Computer visionPyTorchScientific PythonRust

Experience

  1. — present

    inDrive

    Senior Data Scientist · Computer Vision

    Senior since October 2022; previously Data Scientist.

  2. —

    Kantar

    Programmer Analyst → Senior Programmer Analyst

    Data science for advertising and marketing analytics.

  3. —

    VNIIEM Corporation

    Orbital Mechanics Engineer → Space Systems Software Engineer

    Space ballistics, then systems software for satellite onboard electronics.

  4. —

    Medico-mathematical laboratory

    Technician / Researcher · part-time

Full experience on LinkedIn

Open source / Since 2014

Fedora Project

Remote

Package maintenance and scientific software builds.

MaximaFriCAS

Consulting / Since Oct 2024

AAI Scientific Cultural Services

Nicosia, Cyprus · Remote

Part-time research collaboration with a university professor on EEG analysis software in C++ and Qt.

Education / 2010–2015

Applied Mathematics & Computer Science

Specialist degree
Moscow State University of Transport (MIIT)

03 / Elsewhere

Have an interesting problem?

I’d love to hear about it.