Andrei Goponenko

Clearlaw project / AI - based contract-review and management / https://www.clearlaw.ai Achieved: - Developed an algorithmic part of system to help legal teams exchange and analyze documents at fast pace: it included classification models for short and long texts, sequential text classification, NER. - Took part in deployment trained models to microservice infrastructure. - Training language models (GloVe, fastText, ELMo) on domain-specific texts, fine-tuning pretrained models in downstream tasks. - Implemented and deployed brat-based annotation pipeline for text classification task with auto-tagging. - Ad-hoc analysis and EDA for business hypotheses. Technology Stack: Python (Pandas, Numpy, Scipy, Scikit-learn, XGBoost, Lightgbm, Catboost, PyTorch, Keras, Spacy, NLTK, Matplotlib, Seaborn, Plotly, Jupyter, PySpark, Transformers, Hugginface, StanfordNLP, AllenNLP, Azure, OpenAI, Anthropic), Scala, Linux, Git, AWS, Docker, SQLLite, Brat.

Korn Ferry project / Recommendation platform for HR specialists / https://kornferry.com Achieved: - Developed ML and service parts of graph for career transitions which are used in recommendations. - Created service to extract candidate skills from the raw text. Technology Stack: Python (Pandas, Numpy, Scikit-learn, spaCy, Jupyter), Azure (Databricks, Pipelines, Devops), PySpark, Git, YAML.

Minnow project / Aggregator for series and movies from the most popular streaming platforms / https://minnowtv.com Achieved: - Developed matching algorithm for links from streaming platforms and corresponding movies / shows on Minnow's site: brought matching rate 1.7x higher for movies and 2.1x higher for shows. Matching algorithm included more thorough data preprocessing, search by several similarity metrics, fuzzy matching by several fields, validation top-k candidates. - Trained language models (FastText) on domain-specific texts. - Worked on data analysis to enhance overall data quality. Technology Stack: Python (Pandas, Numpy, Scikit-learn, FastText, Plotly, Jupyter), PostgreSQL.

Anagram project / Insurance billing automation SaaS for doctors and patients / https://www.anagram.care Achieved: - Built predictive models for set of benefits based on prior data about patients: list of benefits and coverage, copay, quantity for each benefit. - Cluster analysis of insurance plans for better understanding of patient segments being verified in Anagram app. - Built churn prediction model and tested the bunch of hypotheses for key components of customer health scoring system. Technology Stack: Python (Pandas, Numpy, Scipy, Scikit-learn, imbalanced-learn, XGBoost, Lightgbm, Catboost, UMAP, SHAP, LIME, Matplotlib, Seaborn, Plotly, Jupyter), Linux, Git.

Clearlaw project / AI - based contract-review and management / https://www.clearlaw.ai Achieved: - Developed an algorithmic part of system to help legal teams exchange and analyze documents at fast pace: it included classification models for short and long texts, sequential text classification, NER. - Took part in deployment trained models to microservice infrastructure. - Training language models (GloVe, fastText, ELMo) on domain-specific texts, fine-tuning pretrained models in downstream tasks. - Implemented and deployed brat-based annotation pipeline for text classification task with auto-tagging. - Ad-hoc analysis and EDA for business hypotheses. Technology Stack: Python (Pandas, Numpy, Scipy, Scikit-learn, XGBoost, Lightgbm, Catboost, PyTorch, Keras, Spacy, NLTK, Matplotlib, Seaborn, Plotly, Jupyter), Linux, Git, AWS, Docker, SQLLite, Brat.