Hi, I'm Andrei!
👋 About Me Hi, I’m a Machine Learning Engineer passionate about embedded AI and real-world applications of neural networks.
Currently, I work at Polyn Technology, where I focus on deploying and optimizing neural networks for edge computing and industrial microchips. My work involves:
-
Model quantization and optimization for ultra-low-power hardware
-
Research on neural networks for friction coefficient estimation in autonomous systems
-
Porting and adapting ML models for tiny, efficient deployment in industrial contexts
Previously, I was part of the Sber AI Lab, where I contributed to the development and refinement of LLaMA-based language models and built robust ML pipelines in a production-scale environment.
🧠 Interests & Activities
-
🔬 Research in embedded neural inference and sensor-based learning
-
🤖 LLMs, edge AI, and hybrid RAG systems
-
🎓 Member of the REU DS Club
-
🏕 Participated in the SMILES Summer School by Skoltech
-
🎉 Active attendee of Data Science and AI festivals and hackathons
More stuff about me
-
Drawing up an investor's risk profile for his transactions. Using the LightGBM algorithm on aggregated data in conjunction with the deep learning LSTM model, which predicted the investor's class by the sequence of his transactions. We used the PCA algorithm, feature engineering. It turned out to raise the F1 score from 0.4 to 0.65
-
In the framework of the project-generation of maps with the help of the VAE and the subsequent DCGAN.Creating images with a dimension of 64x64 pixels. the modified VAE made blurry pictures, so it was decided to use deep convolution GAN, which already created clear pictures. All architectures were written from scratch.
-
Sugestive servise is an algorithm that offers auto-completion based on the entered data. Using the trie and reversed trie algorithms.
-
Creating product embeddings. Detection of anomalies in transactions. ETL hackathon solution from Glow byte using pyspark and airflow.
Coding: Python (Pandas, Numpy, Matplotlib, Sklearn, catboost, xgboost, LightGBM, PyTorch, Seaborn, PySpark, transformers, Optuna, ML flow, FastApi, DVC), SQL, HTML/CSS, Git, Docker, Airflow,
Theory: Statistical, Data analysis, ML algorithms (Gradient boosting, Random forest, Regressions, Clustering, Anomaly Detection, etc.), Sequential Neural Networks, Convolution Neural Networks, NLP, GAN, Attention Mechanism, RecSys
