NLP developer / Research / Data Scientist Specialist with 3 years of experience, I can train bert-like, LLM models and work with multi-agent systems and RAG. I have participated in 25+ hackathons, like to travel the World and solve uncommon problems quickly and efficiently. In my free time I am engaged in mentoring and manager of AI Knowledge club in NUST MISIS.
📫 How to reach me: danilamalinin1@gmail.com
- NUST MISIS (Since 2021)
- Deep Learning School [2020-2021]
- Tinkoff Education ML/DL [2020-2021] (certificate)
- Yandex_lyceum [2019-2021] (certificate)
- Yandex_lyceum++ (ML) [2020-2021] (certificate)
- Введение в Data Science и машинное обучение [stepik]
- Hackathon Nornickel [2024] (2/40 🥈) - We provided an architectural solution for searching through Word and PDF files. We used ColPali, the Qdrant database, YOLOv10 to detect text, images, and Qwen with Vikhr-Nemo to answer the question.
- Alfa hack [2024] (5/50) - Implemented Auto ML on alfa bank data with the best quality on 15 datasets.
- Tatneft Hackathon [2024] (3/40 🥉) - Created an analog of Perplexity based on own parser and architecture. Provided a fast SOTA RAG system on TATNEFT documents.
- X5 Tech Hack [2024] (5/30) - Pipeline implementation based on DeBERT, NER and llama-3 to solve the problem of hallucination classification in LLM. F1 score = 0.95.
- Tender-hack [2024] (finalist) [NLP] - the case is "Creating a functional for generating names and characteristics STE in a universal form", where we preprocessed data by nltk, implemented vector database FAISS with embeddings of sentense transformer LaBSE (to work with russian tokens), WildBerries parser, word validation based on n-gramms, t9, pyspellchecker
- Gagarin.Hack [2024] (4/30) - Predicting the sentimentality of news (based on company) by NER and rubert-tiny
- Purple Hack [2024] (3/40 🥉) - 3rd place in the competition from MIPT and sber in the case "Forecasting the outflow of FL payroll clients", where we worked with features and data analytics [ML]
- Talent Match [2024] (2/56 🥈) - 2 place of competition from the company "SENSE Group" supported MSU, with the task of "Сomparison of resumes and vacancies" [NLP]
- AI Energy [2023] (3/25 🥉)- 3 place of international competition with the task of "Disc-type glass insulators absence detection"
- Yandex_cup [2023] (20/236) - Хакатон от Yandex, трек: распознавание жанра музки по эмбедингам (разного входного размера). Задача: MultiLabel Classification
- IVision_hack [2021] (1/30) - Хакатон от Кванториума г. Петрозаводск, где я решил задачу BinaryClassification для компании "Ситилинк" и сделал полноценную модель для распознования оттока клиентов (0 - уйдет, 1 - останется)
- Yandex_alice_hackathon [2021] (30 место) - Навык для голосового ассистента Алиса, написан на языке Python как проект для конкурса "Премия Алисы". Аналог игры Манчкин, только с ограничым, но с таким же интересным арсеналом🔥
- Search_for_superheroes [2021] - Поисковая система, реализованная, как интерфейс настоящей Яндекс строки. Выдача ссылок на странички с описанием супер геровев по запросу, который проходит оработку и tf-idf предсказание по косинусовой близости (Проект Тинькофф поколения)
- Yandex_Race [2021] - Полноценная игра на фреймворке PyGame, где необходимо зарабатывать деньги, принимать заказы и весело разъезжать по карте, покупая новые машины; (Проект Яндекс Лицея)
- Programming: Python, C++
- ML: Scikit_learn, Xgboost/Catboost
- Data Workflow: Pandas, SQL, Numpy, PySpark
- Deep Learning: PyTorch, OpenCV
- Deployment: Git
- Extra: PyGame, PyQt5
- Kaggle Expert striving to achieve Kaggle Grandmaster
- Manager of the AI knowledge club in MISIS.
- Creator and author of Swagman_trip telegram channel. Subscribe to recieve messages about my open-source activities.
- Currently studing in MISIS. Bachelor of Science. Applied mathematics.
- Do sports, in free time play volleyball.
- English: Intermediate (B2)
- Russian: Native
- French: Elementary (A1)