Jul 2021 – Mar 2024 · 2 yrs 8 mos · Berlin, Germany · On-site

• Conundrum is an Industrial AI company that develops a SaaS platform for optimal control of metals and mining assets.
• 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀:
• 🟢 Led a team of 7 data scientists in the development of machine learning modeling and optimization solutions.
• 🟢 Covered client communication, team management, problem formulation, model development, and solution deployment of complex ML / optimization solutions.
• 🟢 Data Science team hiring, growth, and development.
• 🟢 Support of the sales and business development from the AI-product strategy and pre-sale perspectives.
• 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗺𝗲𝗻𝘁𝘀:
• ✅ Increased by $60M+ customer revenue by delivering 4 real-time end-to-end ML solutions that optimized ore processing and smelting plants. The solutions showed 2.5%+ increase of plants throughput and stabilized by 25%+ of the product variability.
• ✅ Identified potential $50M customer revenue improvement by completing 5 bottleneck analysis projects for Tier-1 metals and mining companies, resulting in new projects with them.
• ✅ Grew and structured a DS team from 5 to 20 data scientists who formed 4 internal teams responsible for specific solution directions.
• ✅ Supported closing of 3 multi-million contracts by providing technical support to the sales and business development teams.
• ✅ Reduced by 400%+ the time of development and deployment of ML applications by creating a machine learning pipeline template based on Kedro framework.
• ✅ Decreased from 15 to 3 days the data quality assessment time by creating a standard library that generated consistent data quality reports for new client data.
• ✅ Improved by 70% the DS team task completion rate within a sprint by introducing and adapting Scrum practices.
• 𝗨𝘀𝗲𝗱 𝘁𝗼𝗼𝗹𝘀 𝗮𝗻𝗱 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀:
• Git | GitLab | Kedro | Great Expectations | MLFlow | Scikit-Learn | XGBoost | CatBoost | PyTorch | Keras | Numpy | Pandas | SHAP | Plotly | Matpotlib | Seaborn | Scipy

Feb 2021 – Jul 2021 · 5 mos

• Ctrl2Go is one of the major AI solution providers for the railway and energy industries in the RCIS region.
• 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀:
• 🟢 Led a team of 5 data scientists in the development of predictive monitoring machine learning models for various industrial equipment.
• 🟢 Structured and optimized the company's business to enhance the efficiency of the data science team's contributions.
• 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗺𝗲𝗻𝘁𝘀:
• ✅ Increased by 2.3% smelting plant production by supervising the solution development for optimal ore blending using MINLP formulation in Pyomo.
• ✅ Created a comprehensive model database/library for quicker deployment of the existing solutions in new projects.
• ✅ Formulated product development directions to the C-level management of the company based on the existing clients and partners and the data science team's expertise. It allowed the company management to prioritize development directions, understand risks, and estimate costs.
• 𝗨𝘀𝗲𝗱 𝘁𝗼𝗼𝗹𝘀 𝗮𝗻𝗱 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀:
• Git | GitLab | Kedro | Great Expectations | MLFlow | Scikit-Learn | XGBoost | CatBoost | PyTorch | Keras | Numpy | Pandas | SHAP | Plotly | Matpotlib | Seaborn | Scipy | Pyomo

Oct 2019 – Feb 2021 · 1 yr 4 mos

• Zyfra is one of the major Industrial AI solution providers in the RCIS region.
• 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀:
• 🟢 Led a team of 3 data scientists in the development of ML solutions for problems in the petroleum industry including process modeling, optimization, and anomaly detection.
• 🟢 Covered Industrial AI-based product strategy formulation, management, development, and sales.
• 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗺𝗲𝗻𝘁𝘀:
• ✅ Increased by 1.2% oil and gas production for 2 clients by leading the from-scratch development of the
• ”AI Oil Optimization Assistant” solution. This is web-based ML software for accurate prediction and optimization of oil and gas production in real-time by using machine learning and advanced dynamic optimization algorithms.
• ✅ Increased by 5% the drilling speed and reduced by 50% the stuck pipe occurrence on a pilot test by developing a real-time ML solution to detect drilling pipe sticking events.
• 𝗨𝘀𝗲𝗱 𝘁𝗼𝗼𝗹𝘀 𝗮𝗻𝗱 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀:
• Git | GitLab | Kedro | Great Expectations | MLFlow | Scikit-Learn | XGBoost | CatBoost | PyTorch | Keras | Numpy | Pandas | SHAP | Plotly | Matpotlib | Seaborn | Scipy | Casadi

Aug 2017 – Aug 2020 · 3 yrs · Trondheim, Trøndelag, Norway

• I was developing advanced industrial ML approaches for combining machine learning with process engineering physics for multiphase flow and reservoir modeling and equipment degradation monitoring.
• 𝗔𝗰𝗵𝗶𝗲𝘃𝗲𝗺𝗲𝗻𝘁𝘀:
• ✅ Reached TOP 6 most cited papers in Computers and Chemicals Engineering Journal (Q1) within
• the last 3 years based on the Scopus citation system.
• ✅ Reached TOP 3 in the ranking of most downloaded papers for 2 publications in 2 Q1 journals.
• ✅ Reached over 300 citations within 3 years of active publishing.
• 𝗨𝘀𝗲𝗱 𝘁𝗼𝗼𝗹𝘀 𝗮𝗻𝗱 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀:
• Scikit-Learn | XGBoost | PyTorch | Keras | Numpy | Pandas | Plotly | Matpotlib | Seaborn | Linear Regression | Bayesian Neural Networks | Feed-Forward NN | Gradient Boosting | Random Forest

Jan 2017 – Jul 2017 · 6 mos · Moscow, Moscow City, Russia

• ✅ Developed a model to predict severe slugging occurrence in pipeline networks and stability map representation.
• ✅ Developed a model to calculate transport of solids (flow regime and pressure drop) in single and multiphase steady-state flows for different flow patterns.
• ✅ Adopted industrial standards to calculate erosion rate in pipelines with single and multiphase flows.
• ✅ Developed a model of reciprocating compressors for equipment selection and performance evaluation purposes.

Dec 2016 – Jul 2017 · 7 mos · Moscow, Moscow City, Russia

• ✅ Developed a calculation spreadsheet for the estimation of an oil production increase by wellhead compression technology.
• ✅ In a team of 2 application and 5 service engineers, completed 6 projects on supplying reciprocating compressor parts according to the customers’ needs.
• ✅ Assisted my colleague in a project of reciprocating compressor upgrade and revamp by developing a 3D compressor model in SolidWorks software.

Jun 2015 – Jul 2015 · 1 mo · Delft, South Holland, Netherlands

• ✅ Developed a dynamic model of a reciprocating compressor for CO2 compression in Modelica.