Maksim Shaposhnikov Email and Phone Number

Worked in the Machine Learning Laboratory. Applying ML in FinTech.Applying Deep Learning to Credit Loan Applications.Using RNNs/Transformers on transactional data of the bank’s clients to compute credit scores for the loan applicants.Produced significant financial gains (+1.5 Gini uplift over all segments) both in production and offline experiments.Technologies: Hadoop, PySpark, MS SQL, PyTorch, TensorFlow, Airflow.Contributed to several media events.1) Co-host of the Machine Learning competition Alfa Battle 2.0. Link: https://boosters.pro/championship/alfabattle2/overview2) Co-host the summer educational program for high school students. Link: https://habr.com/ru/company/alfa/blog/569208/

Instructor on the course "Object detection for food on tray recognition task".Taught schoolchildren (9-10 grades) deep learning and CNN's for object detection problems. The result of the project was the mobile app that takes a photo of the tray with food at the canteen, sends an image to a server, detects all dishes, and accepts payment for an order.More details about the project and my students are available further via the link. Technical details (dataset collection, neural network architecture, backend, and mobile app ) are explained in the blog post written by the head of ML laboratory at Alfa-Bank Evgeny Smirnov.

Worked in a RecSys team.Applied Deep Learning for the Smart NewsFeed ranking (sorting posts by user preferences).Implemented DSSM alike architecture for pairwise ranking.Improved target offline metric (grouped ROC-AUC) by a significant margin (+0.03 increase).Technologies: Hadoop, PySpark, PyTorch, Tensorboard.

Worked in a voice assistant team. Designed a context-based dialogue platform for conversational skills.Improved target offline metric (macro-F1) for banking classifier using the ULMFIT model by 1.97. Technologies: ElasticSearch, Keras, Tensorflow.

Worked in a RecSyS team.Developed the recommendation system for the Q&A web service for bookkeepers. Reproduced the DSSM architecture for detecting similar questions for the user query. Applied topic modeling and clustering algorithms for automatic content summarization and annotation. Technologies: Python, sklearn, gensim, BigARTM, Tensorflow, implicit, SQL.

Worked in an NLP team.Developed a content-based classifier to detect pages of different categories: media, online-market, blog, forum. Developed adult content web-page classifier to filter pages from the monitoring platform. Technologies: selenium, asyncio, sklearn, lightGBM, xgboost, vowpal wabbit.