— RAG Expertise: I have extensive experience in building advanced RAG systems, with a strong background in full-text and semantic search technologies. My work includes a robust machine learning stack, featuring LightGBM, SBERT embeddings, Rerankers, and LLMs, implemented using both proprietary and open-source models like OpenAI, Mistral, Anthropic, and Llama.— Benchmarking Achievements: I developed an advanced RAG solution that outperformed LangChain on the FinanceBench benchmark, improving baseline accuracy from 19% to 53.3%.— End-to-End Project Expertise: In my recent end-to-end projects, I enhanced the Askrobot system, achieving a consistent 96% accuracy rate in AI-driven responses, providing reliable support for both customers and employees.— Product Development: I have developed and launched multiple AI-powered Telegram bots that serve Russian-speaking immigrant communities across various countries.

— Developed a Telegram bot (https://t.me/AskIsraelBot) that uses OpenAI's ChatGPT and relevant content from popular Russian-speaking immigrant telegram chats to provide accurate responses;— Developed a Telegram bot (https://t.me/EduSubtitlesBot) capable of generating dual-language subtitles for YouTube videos, preserving the original language while adding a Russian translation.— Launched several Telegram channels (https://t.me/hebrew_trainer, https://t.me/hebrew_trainer_numbers, https://t.me/hebrew_videos) aimed at helping Russian-speaking immigrants in Israel learn Hebrew, with a focus on improving reading and number listening skills;Utilized the following stack to build the projects:— Frontend: React JS;— Backend: Python, PHP, Google Cloud Functions, ElasticSearch, Telegram API, OpenAI API;— Pictures and video automation: Remotion;

— Designed and implemented a data automation system to collect trading data for top 500 S&P and top 100 NASDAQ stocks, ForEx, and Crypto pairs using a crawler to fetch trading, short, and fundamental data from APIs in 30-minute intervals;— Created a MySQL database to store collected data and calculated trading indicators, to extract trading signals and to automate the process of searching for MACD divergences;— Built training sets based on the financial result of trading signals and developed a TensorFlow model for training and predicting trading signals;— Created a Telegram bot (https://t.me/TradeWizardBot) capable of displaying price charts and technical indicators for stocks, forex, or crypto pairs, as well as publishing stock financial data;— Launched a Telegram channel (https://t.me/MACD_Divergences) to broadcast MACD divergence signals;— Launched a YouTube (https://www.youtube.com/@TradeWizardBot) channel featuring educational videos on practicing MACD divergences;Utilized the following stack to build the service:— Frontend: React JS, Bootstrap, Vue;— Backend: PHP, MySQL (20+ tables, triggers, procedures, functions), Python, Google Cloud Stack (App Engine, Cloud Functions), Telegram API;— Video automation: Remotion;— Workflow automation: Jenkins;— ML: TensorFlow, Google Colab Pro, performed tests on an Apple M1 environment to enhance the training speed of the system;

Developed an innovative platform to streamline blue-collar outstaffing companies, including moving companies:— Implemented automated search for new employees, using robo-calls, SMS messages (for older phones), and push notifications to gather responses;— Developed an automated search for half-time or irregular jobs, utilizing a database of registered employees and a search functionality similar to Uber;— Created time scheduling capabilities for full-time jobs, including geo- and face-check-in to manage and monitor employees, with comprehensive settings to calculate wages for workers and dispatchers.Utilized the following stack to build the platform:— Frontend: React JS, Bootstrap, React Native, WebSocket (Pusher.js);— Backend: PHP Yii2, MySQL, Google Cloud Stack (App Engine, Cloud Functions, Datastore (noSQL), CDN Storage, Message Queue), Yandex Cloud (similar to AWS: MongoDB, Message Queue), OneSignal for web and native push notifications, TensorFlow for face-check-in fraud detection;— DevOps: Jenkins, Selenium tools, Katalon IDE;

Subcast helped to improve listening comprehension of English speech in just a few months. The app featured many American and English podcasts on various topics and difficulty levels.— Led the development and launch of a language-learning app that utilizes speech-to-text and lexical analysis to help users improve their English listening comprehension;— Developed the app's front-end using React Native and app’s landing page using React JS;— Integrated speech-to-text system that utilized IBM STT API (and later Mozilla Deepspeech) to synchronize podcast audio and transcript;— Integrated several translation APIs, including Google, Prompt, Yandex, Baidu, Pearson, Jisho, Bluedic and Glosbe.— Integrated Stanford Lexical Parser to enhance the app's language translations;— Implemented core functions for audio file management using Objective C and Java, and developed an audio fingerprinting system to dynamically synchronize audio segments, accounting for changes in audio content, such as the addition of advertisements during streaming services;Utilized the following stack to build the app:— Frontend: React Native, Objective C, Java, React JS;— Backend: PHP Yii2, MySQL, Google Cloud Stack (App Engine, MySQL);— Workflow automation: Jenkins;

— Launched the online kitchen cabinets editor and gallery, enabling customers to design and order custom cabinets with ease;— Developed an online platform that allowed customers in Saint Petersburg to order custom kitchen cabinets from our co-founder's furniture production facility;— Generated leads from customers outside of Saint Petersburg and selling them to other furniture production facilities;— Conducted two R&D projects utilizing Machine Learning: the first involved creating a set of models capable of generating a kitchen design project in the Umugol editor from a photo, the second focused on classifying photos from AirBnB as either a kitchen or not;Utilized the following stack to build the service:— Frontend: React JS, Bootstrap, Three.js, WebSocket (Pusher.js)— Backend: PHP Yii2, MySQL, Google Cloud Stack (App Engine, Cloud Functions, MySQL, Datastore (noSQL), Message Queue);— DevOps: Jenkins, Selenium tools, Katalon IDE;— ML: TensorFlow, Google Colab Pro, Mask R-CNN, MIDAS, Open3D, Transformers;

— Led a team of data analysts and developers;— Developed a custom Business Intelligence platform to collect and analyze key performance indicators (KPIs) for Auto.ru, Vertical Services, and their competitors. The platform was based on ClickHouse to collect and process raw data from user logs in Yandex.Metrika and advertising network logs in Yandex.Direct;— Launched an A/B testing platform to analyze user and revenue metrics and drive data-driven decisions to improve product features and customer satisfaction;— Executed performance marketing campaigns to optimize customer acquisition and retention rates;

Conducted experiments to integrate specialized search engines into common search results to directly display classified listings, aiming to improve user experience and increase efficiency.Led research and development efforts for a personal assistant based on Yandex services, aimed at enhancing user experience and efficiency. Collaborated with cross-functional teams to identify product features that aligned with business objectives and user needs. While the project was not launched, it involved extensive experimentation and prototyping.— Ran experiments on SERP layout and relevance to optimize user metrics;— Collaborated with engineering, design, and marketing teams to develop and launch new SERP products and features;— Conducted market research and user testing to identify and prioritize new SERP features and improvements;— Analyzed performance metrics and reported on KPIs to upper management;

Led the development of specialized search services such as Yandex.Auto, Yandex.Jobs, and Yandex.Realty, along with their corresponding back-office products including a partner dashboard and quality assurance service.Led the development and launch of an aggregated and user-generated reviews service on Yandex.Maps.Led the development and launch of a specialized search engine to search through the apps in AppStore and Google Play.Led the development and launch of multiple projects within Yandex Search, including a "parallel" search for each vertical service, customization of search results for thematic search results, and extending search snippets with data from vertical searches. These projects involved close collaboration with cross-functional teams, including engineering, design, and product management.— Led a team of product managers and developers to develop and launch new products and features;— Managed the product roadmap and prioritized product features based on business objectives and user needs;— Developed and managed budgets, timelines, and resources;— Developed and tracked product metrics to ensure product success and meet KPIs;— Developed product metrics and conducted A/B tests to optimize product performance;— Conducted user research and feedback analysis to inform product development;— Conducted market research and worked with cross-functional teams to identify product opportunities;— Conducted quality assurance on classified listings to ensure high product quality and user satisfaction;

Led the relaunch of Mnogoavto.ru as Yandex.Auto, a specialized search engine that collected all car classifieds across Russia, Belarus, Ukraine, and Kazakhstan.Launched an automotive catalog for Yandex.Auto that included detailed specifications, technical data, and high-quality stock photos.Launched a specialized job search engine that aggregated vacancies from all major job classifieds in Russia, Belarus, Ukraine, and Kazakhstan.— Developed and executed plans to bring new products to market;— Managed a team of engineers to develop new products;— Collaborated with cross-functional teams, including design, marketing, and legal, to successfully launch new products;— Worked closely with executives to define product vision and product strategy;— Established and nurtured partnerships with classifieds to facilitate the operation of specialized search engines;

Developed a specialized search engine that scanned multiple car classifieds. The startup was acquired by Yandex.

Collaborated with a team of software engineers to design, develop, and maintain a web-based application for T-Mobile.co.uk, specifically focusing on the shopping cart and order page.

Co-founded a small web development company with a focus on delivering web-based solutions to small businesses, specifically specializing in developing websites including landings, e-catalogs, e-shops, and other types of web-based solutions.