David Dale

I am an IT specialist, working mostly with natural language processing.
I am from Russia, but now I am living in Paris.
I have a full time job, and I do not want to change it. But I may be open to side projects, if they are interesting and ethical.
I oppose the Russian military aggression and support the independence and territorial integrity of Ukraine.
I support the anti-war and liberation movements in Russian federation with donations, and I may take part in volunteering work for them.

I work in FAIR (Fundamental AI Research at Meta) on NLP (natural language processing) projects:
- No Language Left Behind
- Seamless: Multilingual Expressive and Streaming Speech Translation
As side projects, I develop NLP tools and applications such as:

The first neural machine translation system for the Erzya language (paper, blog post in Russian, code)
A framework for creating chat bots in Python Dialogic
A package for compressing FastText models compress-fasttext
Neural network models for Russian language, such as tiny BERT and Russian T5 in the Huggingface repository
Skills for the Alice voice assistant dialogic.digital

I teach applied mathematics and data analysis, in particular:
- A course on Probability and Statistics for Data Scientists in the Tel Aviv school of Data Analysis (Y-DATA).
- An introduction to Data Science for Products and other managers, a course by MIPT and MathsHub
- Ad-hoc lectures on correct problem statements for data science, natural language processing, basics of mathematical modelling, and other interesting topics.

I worked in the NLP lab at Skoltech university as a research developer
- I worked on problems of text style transfer
- E.g. this paper was accepted at EMNLP
I took part in developing the Alice voice assistant by Yandex:
- Helped creating and improving new scenarios for the Yandex.Station smart speaker
- Boosted the quality and usability of the intent classifier
- Fixed a great deal of bugs (and created some new ones)
I worked as an analyst in Yandex Data Factory
- Helped our clients from manufacturing industries to apply AI for improvement of their production processes
- For example, I helped the Chelpipe plant to harden the pipes faster, without deterioraton in their quality
I helped to manage retail credit risks at Alfa Bank
- Developed new models for forecasting credit losses
- Supported and improved the strategy of issuing credit cards.
- Optimized the algorithm of calling debtors

Github
cointegrated.substack.com: My Substack newsletter "Möbius duct tape" in English
Blogposts on Habr (ru) and on Medium (en)
Google Scholar
Some of my NLP models on Huggingface
My CV
Channels and chats in Telegram
- nlp_jobs - a channel with NLP jobs and résumés
- "Изолента Мёбиуса" - a channel about programming and NLP (in Russian)
- "Матчасть" - a channel about very applied maths (in Russian)
- "Botcamp" - a chat about chat bots and other dialogue technologies (in Russian)
An outdated LiveJournal profile
Youtube recodings:
- My experience building a multiskill voice assistant
- About named entity recognition in Alice (in Russian)
- Basics of feature engineering (in Russian)
- Матчасть 2016: 30 hours of introductory lectures about applied maths and data analysis

Разрабатывал Яндекс.Алису:
- Помогал создавать и улучшать новые сценарии для Яндекс.Станции
- Прокачивал классификатор интентов
- Фиксил баги и создавал новые
Работал аналитиком в Yandex Data Factory
- Помогал нашим клиентам из промышленности применять ИИ для оптимизации процессов
- Например, помог ЧТПЗ быстрее закаливать трубы, не ухудшая их качество.
Работал в Управлении розничных рисков в Альфа-Банке
- Строил новые модели для прогнозирования кредитных потерь
- Поддерживал и улучшал стратегию выдачи кредитных карт
- Оптимизировал алгоритм обзвона должников

URL: https://daviddale.me