Mark Best, Machine Learning Developer in Frome, United Kingdom
Mark Best

Machine Learning Developer in Frome, United Kingdom

Member since March 28, 2016
Mark is truly engrossed in his work of studying people and their behaviors. For the past decade, he's worked on automated trading and time series forecasting—gaining hands-on experience with searching for patterns and opportunities in huge data sets. His statistics background enables him to work in the machine learning space building NLP models for various apps. Mark enjoys working with different data sets and extracting unusual insights from them.
Mark is now available for hire




Frome, United Kingdom



Preferred Environment

Git, Eclipse, Windows, Linux

The most amazing...

...thing I've built is a self-tuning automated pricing system for government bond auctions. Clients receive a price based on market conditions and their behavior.


  • Consultant

    2017 - PRESENT
    • Built an HFT cryptocurrency trading platform for multiple different exchanges. The strategies were able to manage risk and orders with low latency even in very difficult market conditions.
    • Implemented Black Litterman research papers for portfolio optimization.
    • Built out a crypto backtesting framework on top of Backtrader to test strategies on Poloniex.
    Technologies: Python 3, Machine Learning, Portfolio Analytics
  • Quantitive Researcher | Trader

    2014 - PRESENT
    Private Trading Company
    • Researched and traded black and grey box crypto and forex trading strategies.
    • Built an algorithmic trading framework for evaluating and productizing strategies; the strategies use a range of statistical and machine learning techniques to find and exploit price inefficiencies.
    • Designed API libraries for data collection, cleaning, and processing of market data.
    Technologies: Forecasting, C++, R, Python, Cryptocurrency, Cryptocurrency APIs
  • Execution Quantitative Researcher

    2020 - 2020
    Deep Grey Research
    • Built HFT execution models for Eurodollar pro-rata books.
    • Implemented C++ versions of Python research code for production deployment.
    • Built a data framework for backtesting strategies on the Eurodollar market in CME.
    Technologies: Python 3, C++, Machine Learning, Trading, CME, Git
  • Quantitaive Programmer | Researcher

    2016 - 2016
    • Built a framework for connecting and storing tick data from CME and Eurex in a custom high-performance database.
    • Programmed a high-performance multi-threaded trading simulator in C++ capable of processing up to 5 million messages per second on a single thread.
    • Designed and integrated Python research tools with the C++ simulator to test trading strategies.
    • Traded the finalized algorithms into a third-party trading platform for the live execution of the algorithm in Eurex.
    Technologies: Agile Software Development, Forecasting, Boost.python, C++, Python
  • Quantitative Analyst

    2013 - 2014
    Credit Suisse
    • Built a toolset for the analysis of client trading performance (trading analytics).
    • Delivered the results via a Tableau framework allowing for the easy distribution, modification, and extension of the analysis.
    Technologies: Agile Software Development, Forecasting, Tableau, R, C++, Python
  • Associate Director

    2012 - 2012
    Eladian Partners
    • Traded and researched an aggressive cross-asset, high-frequency strategy designed for global futures markets (cross-asset futures trading). The strategy used genetic programming to combine and calibrate various alphas. This also included research for a passive variant of the strategy for fixed income market making.
    • Traded and researched a statistical arbitrage and market-making strategy for government bond futures on CME and Eurex.
    • Managed the operations and risk of the strategy as well as built and calibrated a fully functional passive simulator to test improvements to the strategy.
    Technologies: Agile Software Development, Git, C++, Python
  • Quantitative Researcher

    2009 - 2011
    • Developed a pricing model for the European government bond business. The model worked by risk factor decomposition, forecasting, and recomposition to generate far better prices than models used pre-2008.
    • Designed and implemented a probabilistic market making spread model to optimize the P&L, market risk, and balance sheet usage of a government request for a quote (RFQ) bond business. The model allows the probability of trading to be implied from the market given a set of attributes.
    • Researched and traded a high-frequency market making strategy for bond futures on Eurex (German bond future market making). The research included the development and calibration of a backtesting environment as well as deriving and testing trading signals.
    • Built the internal matching engine for the City Velocity IRS business. This is more complex than a futures matching engine since the consistency of spread and butterfly books also needed to be ensured.
    Technologies: Spring, R, Java
  • Master of Science Candidate in Quantitative Finance

    2008 - 2009
    Cass Business School
    • Studied econometrics; built studies of statistical modeling, prediction, and forecasting.
    • Studied computational statistics: the application of computer option pricing models, Monte Carlo simulations, and other numerical methods.
    Technologies: LaTeX, R, MATLAB
  • Quantitative Programmer

    2006 - 2008
    Deutsche Bank
    • Implemented components within EMMA (electronic market making algorithm) to analyze recent client positions to forecast market movements and build positions passively via market making.
    • Designed and built trading components for vanilla IRS, curve spreads and butterflies on LiquidityHub, TradeWeb, and ReutersSwap electronic platforms.
    • Refactored and improved the programming frameworks for distribution of bond prices to third-party platforms such as Bloomberg and Reuters.
    Technologies: Java
  • Performance Analyst

    2005 - 2006
    Deutsche Bank
    • Managed a team of consultants to improve usability and performance of the Paragon Credit Risk system.
    Technologies: Java


  • Low Latency Messaging Platform for Crypto Currency Trading

    Rust is an interesting language and one that I think will improve the algorithmic trading space. Most high-frequency trading platforms are low latency asynchronous messaging systems. Rust's type safety and fearless concurrency make building high-performance, highly concurrent, multi-threaded messaging systems without much pain in alternative languages such as C++. Also, given the lack of run time and garbage collector, the system's latency is not only low but doesn't suffer from latency spikes caused by the GC. This was a great project for Rust and one that has really made me fall in love with it.


  • Paradigms

    Data Science, Test-driven Development (TDD), Agile Software Development
  • Other

    Machine Learning, Forecasting, Simulation Engines, Finance, Econometrics, Communication, Boost.python, Computer Science, Trading, CME, Cryptocurrency, Cryptocurrency APIs, Portfolio Analytics, Algorithmic Trading, Backtesting Trading Strategies
  • Languages

    Python, SQL, Python 3, Java, C++, R, Rust
  • Libraries/APIs

  • Tools

    PyCharm, Eclipse IDE, Git, Subversion (SVN), Excel 2007, Jupyter, MATLAB, LaTeX, Tableau
  • Platforms

    Linux, Eclipse, Windows, Oracle
  • Storage

    MySQL, Kdb+, MongoDB
  • Frameworks

    Flask, Spring


  • Master of Science Degree in Quantitative Finance
    2008 - 2009
    Cass Business School - London, UK
  • Bachelor of Science Degree in Computer Science
    2001 - 2004
    Warwick University - Warwick, UK

To view more profiles

Join Toptal
Share it with others