Aleksei Shipulin

Aleksei Shipulin

UX & Human-AI Interaction Researcher
Paris, France.

About

UX Researcher with 10 years of experience and a background in psychology, currently leading Yandex Search to integrate LLM functionality.

Work

YANDEX
|

Lead UX Researcher

Summary

Over the last two years, my primary focus has been on exploring how LLMs can change the way people retrieve information and on integrating these capabilities into Yandex Search for our 100+ million monthly active users in a clean and helpful way. As the lead of the Search Engine UX team, I ensure that entire Yandex Search remains intuitive and easy to use. This involves negotiating with stakeholders, creating prototypes, conducting qualitative and quantitative UX research, delegating tasks, analyzing data, and assessing the impact of final results on functionality. Part of my work involves transforming classical UX Research workflows with LLMS: RAG-based insights database, qualitative analysis of interview transcripts with long-context models, analyzing open-ended answers in the questionnaires, giving lectures and workshops.

Highlights

Neuro Search: A search engine enhanced by LLM that combines classical web and image search algorithms with a proprietary YandexGPT model to provide direct answers to users' questions. Launched as a separate tab in April 2024, it was integrated into the main search in October 2024, where it appears under the search results in approximately 10% of all search queries. I oversee the UX for this project, including competitor analysis, identifying user needs, testing, and prototyping. (For example, I made this one, built on the idea of LLM-based search as a discovery platform similar to Pinterest that summarizes articles or videos and allows users to dive deeper into any information.)

Making Alice voice assistant and smart speaker: Provided UX research for the first Russian-speaking voice assistant, including competitor analysis, Wizard of Oz testing, and early prototype testing. Through these efforts, Alice became the most popular voice assistant in Russia with 77+ million MAU and 10 million smart speakers sold. Alice offers support in English, Turkish, Arabic, and Kazakh languages, and the smart speakers are available in Saudi Arabia, Kazakhstan, and additional global markets.

Analyzing open-ended surveys with LLMs: In October 2024, I developed a tool that uses LLMs to analyze open-ended survey responses for UX and MR teams across the company. Since then, it has been used more than 1,000 times, analyzing around 1,000,000 responses and saving approximately 8,227 work hours (or 4 years of full-time work). English version: Telegram bot, GitHub.

Collaboration with image generation team: Collaborated with the YandexArt image generation team to create a methodology for picture assessment. Drawing on my background in psychology, I implemented an Implicit Association Test (IAT) to measure participants' psychophysiological reactions to images generated by various models. Using reaction times as an indirect measure of user attitudes, the training team incorporated this data directly into the RLHF stage. I developed the methodology, created a web application (GitHub) for data collection, and conducted initial statistical testing of our hypotheses. We are currently preparing a research paper based on this work.

USETICS
|

UX Researcher

Summary

Key achievement: Conducted “True Intent Study” research for the most popular Russian mobile operator's website (MTS).

USABILITYLAB
|

UX Researcher Intern

Summary

Introduced to the UX Researcher profession. Significant project: Internet Banking Rank.

Education

Lomonosov Moscow State University

Master's degree

Social Psychology

Languages

English

Fluent

French

Basic

Russian

Native

Skills

R
Cursor AI
Qualitative & Quantitative UX Research
Grounded theory
Figma
Tobii
GraphRAG

Certificates

Data Analysis at R language

Issued By

Summer Camp "Letnyaya Shkola"

Interests

American Literature (XX century)

Thornton Wilder, Kurt Vonnegut, Tomas Pynchon, Donna Tartt, Charles Bukowski.

Videogames

Outer Wilds, Talos Principle, Disco Elysium, Doom Eternal.

Philosophy of Language and Knowledge

Ludwig Wittgenstein, St. Augustine, Lev Vygotsky, Yuri Lotman.

TV Series

Fleabag, Devs, Community, Succession, Euphoria, Severance, BoJack Horseman.

Bosnia & Herzegovina (2024) — Shot on iPhone

Projects

LLM-analyzer of open responses in surveys
LLM-analyzer of open responses in surveys

Summary

A Telegram bot that automatically analyzes open-ended survey responses using AI. It categorizes answers efficiently, saving users 12,505 hours compared to manual processing.

Aleksei Shipulin