Thanks to visit codestin.com
Credit goes to blazerer.com

๐Ÿ‘พ๐Ÿ‘พ๐Ÿ‘พ

Logo

Cloud of competences

View My GitHub Profile

Roman Peshkurov

Software engineer based in London, United Kingdom

Telegram / Email / LinkedIn / Github / PDF


๐Ÿ’พ Technical Experience

SDE @ Wintermute Trading Ltd (May 2022 - June 2025)
Development of server components at HFT company

Details
  • Development of pricing- and accounting- components, which are used for every crypto exchange
  • Predictive models implementation, optimisation and delivery to production. Tangibly affected P&L
  • Market data integrations with several crypto exchanges
  • Participated in regular on call shifts, conducted analysis of anomalies and took part in the interviewing process
  • Tools: C++, Python, PostgreSQL, Internal tools

SDE @ One of FAANG (Dec 2020 - Apr 2022)
Antifraud investigations, training of models and development of server components

Details
  • Worked in the antifraud field, so can't share many details
  • Trained models in ambiguous area
  • Developed data pipelines and implemented business logic around them
  • Performed anomalies investigations
  • Managed bootcampers and interns, conducted interviews
  • Tools: Python, Internal ML libraries, Hive/Spark, MySQL, HG, Internal tools

Technical lead @ Yandex.Auto (Jun 2019 - Nov 2020)
Management of the team responsible for the server-side components of the navigation service

Details
  • Designed and developed server-side applications for the integrations of our service with partners (eg. Remote-Access-To-Car)
  • Introduced dashboards and metrics recalculation processes
  • Leaded development of push-service for proprietary devices
  • Developed Logs-On-Demand system for our devices
  • Participated in interviewing, onboarding, mentorship, project management routines of the team
  • Tools: C++14, PostgreSQL, PlantUML, Java, Yandex-tools

Remote-Access-To-Car Demonstration
We integrated our navigation app with the telematics system, so that allowed our customers to have access to car's features (ignition, locks, etc.) via app. Was responsible for the design and development of server components via management of the team of 4 developers.
Here you can see post about presentation from the conference with a part about our project:


Senior SDE @ Yandex.Translator (May 2018 - Jun 2019)
Training translation models and development of data pipelines

Details
  • Designed and developed end-to-end data aggregation pipeline: from application on device to table into the MR-storage
  • Improved translation quality on Turkish language family via training new types of neural models, waiting for patent
  • Implemented Alternatives-For-Translation backend based on neural language model API
  • Tools: C++14, Python, NLTK, TF, Yandex MapReduce, Transformer RNNs, HG

Alternatives-For-Translation Demonstration
Worked on an optimal extraction of probabilities' values from translation model to allow us to suggest alternative words in the translation.
As a result, we obtained server API for this task, and here you can see example of this tool's usage on en-ru language pair:


Turkish Languages Translator and Yakut Language
Worked on the multi-language translation models with Tranformer RNNs and tried to find the best experiments' setup by preparation of corpora and tuning of models.
In the end, we achieved better quality in comparison to our baselines on the az-ru, kz-ru, uz-ru and other turk languages.
Yakut language is endangered one and our translator was the first registered translation system for it. As this language has high significance for the russian indigenous people, this achievement was reflected in the dedicated documentary:


SDE @ Yandex.Appmetrica (Mar 2015 - May 2018)
Server-side development of real-time analytical service for mobile applications

Details
  • Developed distributed fault-tolerant real-time data pipeline (40ยท10^9 rows/day) as a set of microservices
  • Maintained and optimised self-written backend servers' layer (140ยท10^3 RPS of HTTPS in peak)
  • Developed various utils: internal/external APIs, MR-jobs, monitorings and testing solutions
  • Performed investigations on petabytes of data via MR and ClickHouse SQL
  • Tools: C++14, Boost, Poco, MySQL, ClickHouse (developed in my department), Nginx, Python, Yandex MapReduce, Apache Zookeeper, Git

Logs API Demonstration
Designed and developed the server side of this API to allow customers extracting logs of their applications from our storage.
Here you can see how process of the data download looks like:


SDE @ ModuleWorks 3ax Team (Mar 2012 - Feb 2015)
Development of the solutions for the CAD/CAM software

Details
  • Solving sculpture surface machining and computational geometry problems
  • Implemented library for the toolpath smoothing - based my master's thesis on it
  • Developed and maintained project of Multiaxis Roughing (see below)
  • Tools: C++03, Boost, Computational geometry tools, Svn

Multiaxis Roughing Demonstration
This toolpath generation algorithm was implemented as a combination of approaches between 5ax and 3ax project areas. My aim was to make an actual implementation of an already scoped project and pushing it to the production.
Here is the marketing demonstration of this tool:


๐ŸŽ“ Education

Master Program in CS and DA @ Yandex Data School (Sep 2015 - Jun 2017)


Master Degree in Computer Engineering @ Sevastopol National University of Nuclear Energy and Industry (Sep 2008 - Jan 2014)


๐Ÿ† Achievements


๐Ÿ’ฌ Personal Qualities


๐ŸŽธ Outside Of Work