Iliyan Zarov, Machine Learning Developer in London, United Kingdom
Iliyan Zarov

Machine Learning Developer in London, United Kingdom

Member since January 4, 2017
Iliyan is a machine learning engineer and full-stack developer with a decade of experience. He is a co-founder of an AI health-tech company backed by Europe's most prestigious accelerator. Iliyan was previously lead research engineer in the AI Research team at Babylon Health, building and scaling up the AI doctor to millions of users. Before that, he made a massively scalable distributed computing platform and worked as a quant at a hedge fund.
Iliyan is now available for hire


  • myLevels
    Amazon Web Services (AWS), AWS, Docker, Flask, React Native, JavaScript...
  • Babylon Health
    Amazon Web Services (AWS), AWS, Docker, PyTorch, TensorFlow, Python
  • Evoqus
    Amazon Web Services (AWS), OpenStack, Docker, AWS, Julia, Python, Go, Clojure



London, United Kingdom



Preferred Environment

Git, Emacs, Linux

The most amazing...

...thing I've built is an app that uses machine learning to estimate the impact of food on our bodies in real time.


  • Co-founder | CTO

    2018 - PRESENT
    • Deconstructed blood glucose levels time series into constituent signals using Bayesian machine learning techniques. We can then estimate and assign a score to the individual impact of each food. This appears to have never been done before and we filed a patent on this approach.
    • Built an NLP model to classify foods into food groups (e.g., grains, dairy, meat, etc.).
    • Developed a regression model to estimate the contribution of each food group to a food impact score.
    • Reverse-engineered the protocol for communicating with the continuous glucose sensor over.
    • NFC or Bluetooth and implemented it using Kotlin and Swift.
    • Architected and built our offline-first mobile app in React Native with the help of two front-end engineers who I supervised.
    • Designed a custom synchronization protocol to guarantee that personally identifiable data never leaves the data center in the UK (as opposed to, for example, Google's Firebase).
    • Developed a stateless back end exposing a RESTful API using Python and Flask.
    • Created an automatically generated personalized report using LaTeX and Python's Jinja.
    Technologies: Amazon Web Services (AWS), AWS, Docker, Flask, React Native, JavaScript, PyTorch, Pandas, NumPy, Python
  • AI Research Engineer

    2017 - 2018
    Babylon Health
    • Built Dr. Deep, making it up to 10x faster to run medical diagnosis at the same level of accuracy. Dr Deep is a neural network based universal marginalizer for amortized inference in generative models.
    • Published and presented Dr. Deep at the NeurIPS Workshop on Advances in Approximate Bayesian Inference.
    • Researched and implemented large-scale Bayesian inference algorithms for probabilistic models in Python, C++, TensorFlow, and PyTorch.
    • Productionized the Bayesian inference engine used by Babylon’s AI doctor.
    • Designed the massively scalable cloud architecture which is used to provide the AI doctor to millions of users.
    • Introduced PyTorch and Pyro at Babylon.
    Technologies: Amazon Web Services (AWS), AWS, Docker, PyTorch, TensorFlow, Python
  • Founder

    2015 - 2017
    • Created a cloud-based distributed computing platform from the ground up. The code can be executed with a single command across hundreds of CPUs and GPUs in the cloud.
    • Developed RESTful API back-end services with a custom orchestration layer in Clojure.
    • Developed a Go client for batch and fully interactive jobs. The Go client includes support for attaching console with raw terminal mode, tunneling, and large data files transfer.
    • Designed the back-end services to be stateless and resilient with automatic scale-out.
    • Developed end-to-end tests against AWS and a local OpenStack deployment.
    Technologies: Amazon Web Services (AWS), OpenStack, Docker, AWS, Julia, Python, Go, Clojure
  • Quantitative Developer

    2012 - 2016
    Markham Rae
    • Researched and validated volatility quant trading and portfolio allocation strategies.
    • Built swaptions and exchange-traded options pricing tools.
    • Built tools for scenario analysis of yield curves, volatility cubes and model parameters.
    • Built an exotic FX option pricing-&-scenario analysis tool.
    • Transitioned all of the interest rate applications to a new multi-curve framework.
    Technologies: MATLAB, C#, C++, SciPy, Python
  • Senior Developer

    2008 - 2012
    • Built the aggregated liquidity pool platform and trading front end.
    • Played a leading role in multiple projects on order routing and consolidation of hybrid inter-dealer broker and exchange venues for cross-asset and cross-venue trading.
    • Spearheaded the development of indirect order aggression features for implied spread trading.
    • Led the development of a latency and availability platform enhancement project.
    • Worked in a lead role of the architecture and development of a new exchange connectivity and a new clearing platform.
    • Prototyped a distributed data storage and processing engine on top of Apache Cassandra.
    • Developed an F#-based, domain-specific language and contract specification for a commodity product registry and standardization service.
    Technologies: F#, C#, C++


  • myLevels (Development)

    myLevels uses data from continuous glucose monitors and Bayesian machine learning to model the very personalized impact that food has on people’s bodies. We help you see and understand what foods are best for your unique metabolism. People using myLevels for weight loss have dropped more than 10 kilograms.

  • A Universal Marginalizer for Amortized Inference in Generative Models (Other amazing things)

    We consider the problem of inference in a causal generative model where the set of available observations differs between data instances. We show how combining samples drawn from the graphical model with an appropriate masking function makes it possible to train a single neural network to approximate all the corresponding conditional marginal distributions and thus amortize the cost of inference. We further demonstrate that the efficiency of importance sampling may be improved by basing proposals on the output of the neural network. We also outline how the same network can be used to generate samples from an approximate joint posterior via a chain decomposition of the graph.

  • Evoqus (Development)

    Evoqus is a cloud platform delivering effortless scalability for technical computing in Julia. Code can be executed with a single command across hundreds of CPUs and GPUs in the cloud.

    The stateless back end services are written in Clojure. The platform can be accessed using the CLI, written in Go, or the web app built on React/Om and ClojureScript.

    Evoqus grew out of a side project aiming to make it extremely easy to move the computing from my laptop to the cloud on demand. The details of the underlying cloud instances, storage, and network are transparent to the end user while retaining full control over the workers and the environment.


  • Languages

    Go, Julia, C, C++, Clojure, Python, JavaScript, Kotlin, Swift, Java, F#, Scala, C#
  • Frameworks

    React Native, Flask, ClojureScript, Boost
  • Libraries/APIs

    Pandas, SciPy, TensorFlow, NumPy, Compojure, AWS EC2 API, PyTorch, LSTM, Standard Template Library (STL), SpaCy
  • Tools

    AWS ELB, Amazon Virtual Private Cloud (VPC), Emacs, LaTeX, Git, Amazon SQS, AWS IAM, MATLAB, AWS SES
  • Paradigms

    Functional Programming, Asynchronous Programming, REST, Concurrent Programming, Parallel Computing, High-performance Computing, Distributed Computing, Agile
  • Platforms

    Docker, AWS EC2, Amazon Web Services (AWS), Linux, Ubuntu, CUDA, OpenStack
  • Industry Expertise

    Financial Modeling, Financial Engineering
  • Storage

    AWS S3, PostgreSQL, AWS RDS, Cassandra, MySQL, Microsoft SQL Server
  • Other

    Machine Learning, Recurrent Neural Networks, Neural Networks, Convolutional Neural Networks, Deep Neural Networks, Deep Learning, API Design, Data Analysis, Data Analytics, Quantitative Analysis, Financial Markets, Bayesian Inference & Modeling, Elastic Beanstalk, Natural Language Processing (NLP), Probabilistic Graphical Models, Algorithms, Evolutionary Algorithms, GPU Computing, Optimization, AWS Route 53, Elastic Load Balancers, Mathematical Modeling, Containers, Container Orchestration, Probability Theory, Quantitative Finance, High Availability, Networks, Statistical Modeling, Time Series Analysis, Data Structures, Integration Testing, PaaS, Scalability, Network Programming, Medical Diagnostics, Computer Vision, AWS, Wireless Sensor Networks, log4j, Reinforcement Learning, Deep Reinforcement Learning


  • Master of Science degree in Mathematical Trading and Finance
    2009 - 2011
    Cass Business School - London, UK
  • Bachelor of Science degree in Electrical Engineering and Computer Science
    2004 - 2007
    Jacobs University Bremen - Bremen, Germany

To view more profiles

Join Toptal
Share it with others