Distribution Visualizer ↗
Interactive PDF/CDF explorer for common probability distributions.
JavaScript
Plotly
Statistics
Anastasia Sosnovskikh
Engineering Lead
I turn ML research into systems that work at scale. Eight years in industry — and a Master's in Machine Learning currently in progress — taught me that the hardest part isn't the model. It's everything around it.
My background is a bit unusual: Computer Science meets Psychology and Neuroscience. It shapes how I think about building systems that work for people, not just on benchmarks.
Python
R
SQL
Azure
AWS
NLP
How I work
Everything Around the Model
A model is one component. The work is turning an ambiguous need into a reliable system that people can use and the business can measure.
Business need
Translate ambiguity into a problem the team can solve.
Education
Master of Science in Computer Science — Machine Learning Specialization (In progress)
Georgia Institute of Technology, USA
Bachelor of Science with Honours — Computer Science, Psychology, Neuroscience Summa Cum Laude
University of Pittsburgh, USA
International Study Program — Software Engineering Honours
UNSW Sydney, Australia
Recent Experience
Fortune 200 Manufacturing Company — 7+ years
Engineering Lead
2026 – Now
Digital Team, Automotive OEM
Software Engineer
2024 – 2026
Digital Team, Science and Technology
Software Engineer
2021 – 2024
Research & Innovation Team, Science and Technology
Software Engineer
2019 – 2021
Cross-Functional Digital Innovation Team, Science and Technology
Work
Voice-Driven Wearable Application
ML-powered wearable integrating forecasting software.
Python
Voice AI
NLP Voice Application
Voice-driven app with custom NLP engine.
NLP
Python
System Performance Prediction ↗
85+ pipelines, 2x precision improvement, reduced failure probability to 13% via SANN optimization.
R
Regression
Classification
Property Price Prediction ↗
Kaggle-style regression. 30% improvement over baseline via feature engineering and 15+ tuned models.
R
Regression
Bitcoin Price Forecasting ↗
Time-series with exogenous variables. 15% improvement via cross-validation and ARIMA tuning.
Python
Time Series
Recommendation Engine ↗
A-Priori association rule mining built from scratch — no ML libraries.
Python
Algorithms
Writing
The convenience is real. So is the risk.
The Hidden Cost of AI Assistance: Outsourcing Critical Thinking.
Contact
Open to engineering leadership and technical roles across software, data, and AI. If you're building something interesting, let's talk.
Prefer email?