Timur Khasanov - NLP and Software developer, Tiburon Research

Keep in touch with meI'm using Intch to connect with new people. Use this link to open chat with me via Intch app

Work Background

Tiburon ResearchNLP and Software Developer

Dec. 2018Moscow, RussiaWorked on products for text analytics from answers to open-ended questions in online surveys. Delivered such projects as: - Keyword Clouds: Automatic extraction of key words from a given array of short texts from open-ended answers to a question, adjectives and verbs characterizing them, forming user opinions, and quantifying them; merging similar words based on meaning, spelling, and morphology. Used words and sentences embeddings, dependency trees parsing, part of speech tagging, and named entities recognition. Also developed multilingual support for it with the help of Google Translate API. - ChatGPT interviewer bot: Built the backend side of a chat bot that is generating prompting questions based on a survey's context automating the job of a qualitative research moderator using OpenAI API - Sentiment analysis: A classification system to detect positive/negative/neutral answers or reviews using fine-tuned BERT models on corporate data, increased the F1 metric twofold compared to the pretrained russian model - Text similarity tools: Free-form text labeling by a list of opinions and topics; sorting an array of texts by their similarity to ease manual topic labeling. Used pertained sentence embedding models with fine-tuning on scraped labeled data. - Fuzzy brand labeling: API that receives survey data from an open-ended question, finds clusters of brands or names with different spellings based on Levenshtein distance and heuristics, labels which brands are mentioned in every answer. - Gibberish answer filter: Developed tools that classify answers by containing meaningful information, and flag suspicious users who flood the data. - Excel parsers: Scraped data from thousands of manually labeled tables from Excel files on corporate network drives. - Admin panel: Developed a web app for visualizing the output result of designed algorithms, testing their performance, updating databases manually from a browser, so a manager could edit them as well. Daha az göster

Tiburon ResearchAnalyst

Nov. 2017 - Dec. 2018Moscow, Russia- Survey programming: With existing survey engine implemented complex logic in surveys with different scenarios for different conditions using C# scripts - Market research data analysis: Survey data processing, building SPSS tables, calculating various metrics in Python