Data Scientist with a PhD in Mathematics (RWTH Aachen, Germany) and 9+ years in industry across banking, trading and retail: forecasting, document recognition with vision-language models, recommender systems and production ML. Before that, 10+ years of research in optimization and control (NUS, Singapore; RWTH and FAU, Germany) and 30+ publications.
LCR forecasting30 liquidity components, 60 days ahead, daily in production at a bank
Document recognitionVision-language model with human review and export to client-form fields
4.5M MiFIR transactionsCorrected and bulk re-reported to Deutsche Börse
Sales forecasting2M items and warehouse replenishment for a German retail chain
PhD and 30+ publicationsRWTH Aachen (Germany), grade 1 (highest); papers in SIAM J. Control Optim. and Systems & Control Letters
LCR forecasting (production): automated the daily 60-day forecast of 30 balance-sheet components of the Liquidity Coverage Ratio, replacing a manual process. XGBoost, LightGBM, CatBoost, Ridge and Prophet + Optuna ensemble, walk-forward validation, PSI drift monitoring, scheduled runs, PostgreSQL and Streamlit dashboards.
Document recognition (MVP): extraction of Belarusian passport data into client-form fields with the Qwen3-VL-8B vision-language model (vLLM) — page classification, field extraction, human-in-the-loop review, Excel export. Extended to leasing documents: sole-proprietor registration certificates and state-register extracts, multi-page PDFs, a FastAPI backend and a Streamlit operator UI.
RecSys: built and deployed a recommender system that personalizes financial product and service offers. It analyzes client data and interaction patterns to suggest the most relevant product at the right time, improving customer satisfaction and sales.
Cash Management: built and deployed real-time transaction monitoring across all bank branches in Belarus. It detects discrepancies and potential fraud, enabling efficient cash allocation and liquidity management.
Implemented an automated ML solution for sales and next-best-offer (NBO) forecasting: data collection and consolidation (transactions, promotions, customer feedback, competitor prices, holidays and events), metric design, XGBoost, RandomForest and LightGBM models.
Credit risk (retail): developed a behavioural PD model, deployed to production — a logistic regression scorecard with WoE binning, IV-based feature selection, ROC-AUC/Gini validation and PSI stability monitoring; also developed LGD and CF models (not deployed).
Set up the ML pipeline for the production environment
Developed ETL for AWS services and the enterprise data warehouse (EDWH)
IT vendor delivering projects for banks, trading companies and retailers. Roles: Financial Engineer (2017), Data Analyst (2017–2018), Data Scientist (2018–2019).
Retail (2019): delivered ML sales forecasting and warehouse replenishment optimization for a German retail chain: 2M items, 130 categories, 2K partners, 50K daily sales. Per-category models (regression, XGBoost, RandomForest, LightGBM) and an ML pipeline on Azure Databricks and Spark. The forecast automated ordering, cut replenishment costs and enabled price elasticity measurement.
Banking (2017–2018): analyzed rejected MiFIR reports; designed and implemented PL/SQL routines to correct and bulk re-report ~4.5M transactions not acknowledged by Deutsche Börse during the initial MiFIR phase
Trading (2017): forecasting, valuation and reporting tools for long-term contracts with strong seasonality — Monte Carlo pricing with weather derivatives to account for temperature optionality, analysis of historical data from German weather stations
Trading (2018): built data anomaly detection tools and alerting rules for data issues
Trading (2018): implemented portfolio clustering by market liquidity (number of contracts, order volume, turnover) and procedures to quickly explain P&L swings caused by market index and spread movements
Banking: EMIR Level 3 reporting of transactional, collateral and valuation data with defect and root-cause analysis; SQL routines for top-5 execution venues reporting; remediation concept for MiFIR production data issues
Business analysis: requirements for a consolidated EMIR/MiFIR II reporting system, contribution to FSD/BRD; business requirements for valuation and reporting in a new risk management system
Financial services provider (2018): R Shiny ML tool: EDA, feature selection, hyperparameter tuning, classification and clustering, comparative model reports
Trading (2017): reports on positions, portfolio performance and realized profit (WTD/MTD/YTD); custom packages for historical time series data
Autonomous UAV systems: robust, fault-tolerant control, monitoring and forecasting.
Led requirements specification, feasibility studies, forecasts and development proposals
Developed Ground Control Station algorithms (computational geometry, combinatorial optimization) and a reciprocal collision avoidance algorithm (velocity obstacle) for real-time multi-agent simulation
Built a simulation toolbox for formation control of fixed-wing SB-III UAVs with a user manual (industrial use)
Evaluated transfer learning methods on data from Singapore's Defence Science and Technology Agency
Won a Seed Project Grant (Fast Multiple UAV Flight Generation, 2016); 10+ publications in journals and conference proceedings
DFG project “Optimization and Control on Networks”.
In an interdisciplinary team, developed an algorithm deriving differential equations of switching functions for optimal control laws, applied to gas transport networks
Led exercise classes for “Mathematics I for Mechanical Engineers” and “Optimal Control of PDEs”
Postdoctoral Researcher
Sep 2006 — Sep 2008
Friedrich-Alexander University Erlangen-Nuremberg (FAU) · Erlangen, Germany
DFG project “Optimal Control in Gas Networks”.
Developed mathematical models of controllability and optimality for gas networks and numerical simulation and optimization methods using real E.ON Ruhrgas AG data
Analyzed technical reports to quantify the reliability of the company's products
Projects
Water pump failure prediction
Classifying pumps as functional / needs repair / non-functional to plan maintenance; R, Shiny, parallel processing
Anomaly detection in options markets
A learner that finds logical inconsistencies in option Bid/Ask prices and greeks
Multi-UAV mission planning
Multiple TSP with neighbourhoods, revisit requirements and Dubins vehicle constraints; Matlab, C++
Side projects (hobby)
Lakehouse for a digital bank
Data platform from scratch on open source: Apache Iceberg on S3, Trino, Kafka and Debezium (CDC from two core banking systems with migration reconciliation), dbt and Data Vault 2.0 (PIT, bridge, bitemporal as-of queries), Write-Audit-Publish on Iceberg branches, 25 data-quality checks, PII tokenisation and Trino roles, catalog and lineage (dbt docs), Superset dashboards as code with white-label branding (no fork), a central-bank regulatory report with controls and reproducible filing; the whole stack starts with one command
Contact-center speech analytics
Call transcription (faster-whisper) and speaker diarization (pyannote), analysis by four LLM agents (topic, quality control, compliance, summary), role-based Dash dashboard, FastAPI, PostgreSQL; dashboard and documentation (Jupyter Book) deployed on Railway
Generative AI in ComfyUI
Custom ComfyUI node pack in Python for image and video generation and editing via an OpenAI-compatible LiteLLM gateway; workflow converter and end-to-end tests on ComfyUI with a mock gateway; 326 workflows reviewed, about 120 migrated to the custom nodes
Interactive Data Science book
Course notes for myself and my son built with Jupyter Book: SQL runnable in the browser (PostgreSQL in WebAssembly), own diagrams, homework walkthroughs and self-check exercises
Family tree explorer
Streamlit app for a family lineage (about 170 people): OCR of family-tree photos (Tesseract), kinship graph in SQLite imported from GEDCOM/FamilyEcho, interactive sunburst and treemap views (Plotly)
Publications
2026
On modeling a bank's investment strategy
M. P. Dymkov, E. V. Bratkovski, S. Dymkou
International Scientific Conference on Differential Equations (Erugin Readings 2026), vol. 2, pp. 82–84 (Abstract, in Russian)
2026
Mathematical models for managing financial assets
M. P. Dymkov, S. Dymkou, E. V. Bratkovski, S. P. Makarevich
Belarusian Economic Journal, no. 1, pp. 113–135 (In Russian)
2025
Optimization problems for overdetermined multidimensional discrete systems
M. P. Dymkov, S. Dymkou
International Conference on Electrical and Computer Engineering Researches (ICECER 2025), Antananarivo, Madagascar
2025
Linear quadratic optimization problems for overdetermined multidimensional discrete systems
M. P. Dymkov, S. Dymkou
The Eighth Bogdanov Readings on Differential Equations, Minsk, Belarus
2022
A method for relaxing state constraints in nonsmooth optimal control problems
M. P. Dymkov, S. Dymkou
Journal of the Belarusian State University. Mathematics and Informatics, no. 2, pp. 107–114 · DOI
2019
Distributed UAV Loss Detection and Auto-replacement Protocol with Guaranteed Properties
S. Huang, R. S. H. Teo, J. L. P. Kwan, W. Liu, S. Dymkou
Journal of Intelligent & Robotic Systems, vol. 93, no. 1–2, pp. 303–316 · DOI
2018
Optimization problem for linear discrete 2-D systems with boundary control
M. Dymkov, S. Dymkou
International Scientific Conference Dedicated to the 100th Anniversary of E. A. Barbashin (DSSCO-18), pp. 19–21, Minsk, Belarus
2017
Distributed cooperative UAV loss detection and auto-replacement protocol with guaranteed properties
S. Huang, R. S. H. Teo, W. Liu, S. Dymkou
2017 International Conference on Unmanned Aircraft Systems (ICUAS), pp. 393–401 · DOI
2017
Distributed cooperative collision avoidance control and implementation for multi-unmanned aerial vehicles
S. Huang, R. Swee Huat Teo, W. Liu, S. Dymkou
2017 11th Asian Control Conference (ASCC), pp. 222–227 · DOI
2017
Agent model for multi-UAV control via protocol designs
S. Huang, S. H. R. Teo, W. Liu, S. Dymkou
International Journal of Intelligent Computing and Cybernetics, vol. 10, no. 4, pp. 412–429 · DOI
2017
Decentralized real-time optimal control of a group of objects based on state information of their neighbours
S. Dymkou, U. Simankou, R. S. H. Teo
10th International Workshop on Multidimensional (nD) Systems (nDS-2017)
2016
Mathematical models and algorithms for multi-agent control systems
S. Dymkou, A. Markov, U. Simankou
Doklady BSUIR, no. 6 (100), pp. 67–72
2015
Adaptive decentralized control problem for MAS
S. Dymkou, U. Simankou
3rd Workshop on Research, Education and Development of Unmanned Aerial Systems (RED-UAS 2015), Cancún, Mexico
2014
A net graph setting for the gas-pipelines modeling
S. Dymkou, V. M. Dymkou, M. Dymkov
Scientific Journal of Mathematics Research, vol. 4, no. 4, pp. 26–45
2014
Multifunctional transformation method in flow modeling
S. Dymkou, V. M. Dymkou, M. P. Dymkov
IFAC Proceedings Volumes, vol. 47, no. 3, pp. 7013–7018 · DOI
2014
Boundary control in large scale distributed gas transportation network systems
S. Dymkou, M. Dymkov
Researches and Applications in Economics, vol. 2, no. 2, pp. 21–25
2014
Boundary control problem in distributed transportation networks
S. Dymkou, M. Dymkov
XVI International Scientific Conference on Differential Equations (Erugin Readings 2014), Novopolotsk, Belarus
2013
Multifunctional transformation method in gas pipelines modeling
M. Dymkov, V. M. Dymkou, S. Dymkou
XV International Scientific Conference on Differential Equations (Erugin Readings 2013), vol. 2, pp. 83, Grodno, Belarus (Abstract)
2013
Optimization problems for the gas transportation networks
M. Dymkov, S. Dymkou
VI International Conference “Economic Growth of the Republic of Belarus: Globalization, Innovation, Sustainability”, Minsk, Belarus (Abstract)
2012
Repetitive and 2-D systems theory approach for modeling in gas networks
M. Dymkov, S. Dymkou
2012 IV International Conference “Problems of Cybernetics and Informatics” (PCI), pp. 1–4, Baku, Azerbaijan · DOI
2012
Repetitive models in gas transportation networks
S. Dymkou, M. Dymkov
2012 17th International Conference on Methods & Models in Automation & Robotics (MMAR), pp. 433–438 · DOI
2012
Modeling in gas pipelines networks: Repetitive and 2-D system theory setting
S. Dymkou, M. Dymkov
International Conference on Mechanical Science and Engineering (ICMSE 2012), Beijing, China
2011
Modeling and control of a sorption process using 2D systems theory
M. Dymkov, K. Galkowski, E. Rogers, V. Dymkou, S. Dymkou
The 2011 International Workshop on Multidimensional (nD) Systems, pp. 1–5 · DOI
2010
2D production model with environmental factors
M. Dymkov, A. O. Brilevski, S. Dymkou
III International Conference “Problems of Cybernetics and Informatics” (PCI-2010), pp. 42–45, Baku, Azerbaijan
2008
Constrained Optimal Control Theory for Differential Linear Repetitive Processes
M. Dymkov, E. Rogers, S. Dymkou, K. Galkowski
SIAM Journal on Control and Optimization, vol. 47, no. 1, pp. 396–420 · DOI
2008
Optimal control of wave linear repetitive processes
M. Dymkov, E. Rogers, K. Galkowski, S. Dymkou
Systems & Control Letters, vol. 57, no. 11, pp. 940–945 · DOI
2008
Constrained optimal control theory for differential linear repetitive processes
M. Dymkov, E. Rogers, S. Dymkou, K. Galkowski
2008 47th IEEE Conference on Decision and Control, pp. 263–268, Cancun, Mexico · DOI
2008
Optimal control problem for the pipeline network model
S. Dymkou, G. Leugering
International Conference on Differential Equations and Topology Dedicated to the 100th Anniversary of L. S. Pontryagin, Moscow, Russia
2008
Optimal Control of Non-stationary Differential Linear Repetitive Processes
S. Dymkou, M. Dymkov, E. Rogers, K. Galkowski
Integral Equations and Operator Theory, vol. 60, no. 2, pp. 201–216 · DOI
2008
Boundary control problem for the gas network model
S. Dymkou, G. Leugering
International Conference “Dynamical Systems: Stability, Control, Optimization” (DSSCO-08), Minsk, Belarus
2007
Repetitive processes modelling of gas transport networks
S. Dymkou, G. Leugering, G. Jank
2007 International Workshop on Multidimensional (nD) Systems, pp. 101–108 · DOI
Graph and 2-D systems approach in gas transport network modeling
S. Dymkou, G. Jank, T. Azevedo-Perdicoúlis
International Journal of Tomography and Statistics, vol. 6, pp. 21–27
2006
Graph and 2-D Optimization Theory and their Application for Discrete Simulation of Gas Transportation Networks and Industrial Processes with Repetitive Operations
S. Dymkou
RWTH Aachen University
2006
Graph and 2-D systems approach in gas transport network modeling
S. Dymkou, G. Jank, T. Azevedo-Perdicoúlis
IFAC Workshop on Control Applications of Optimization (CAO'06), pp. 165–171, Cachan, France
2005
An approach to controllability and optimization problems for repetitive processes
S. Dymkou, E. Rogers, K. Galkowski, M. Dymkov, D. H. Owens
International Conference on Stability and Control Processes (SCP-2005), vol. 2, pp. 1504–1516, Saint Petersburg, Russia
2005
2-D control theory in gas transport network modeling
S. Dymkou, G. Jank
The Fourth International Workshop on Multidimensional Systems, 2005. NDS 2005., pp. 241–246 · DOI
2005
2-D theory settings and correct discretization in the gas network simulation
S. Dymkou
International Conference “Erugin Readings X”, pp. 109–110, Mogilev, Belarus (Abstract)
2004
Time optimal control problem for a class of linear differential-algebraic systems with delay
S. Dymkou, G. Jank
16th International Symposium on Mathematical Theory of Networks and Systems (MTNS 2004), Leuven, Belgium
2003
Approximations of the constrained nonsmooth optimal control problems and their applications for the robot planning motion
S. Dymkou
ICIAM 2003 Congress, Sydney, Australia
2003
A delay systems approach to linear differential repetitive processes: Controllability and optimization
S. Dymkou, E. Rogers, M. Dymkov, K. Gałkowski, D. H. Owens
IFAC Proceedings Volumes, vol. 36, no. 19, pp. 333–338 · DOI
2003
Controllability and optimization for differential linear repetitive processes
S. Dymkou, E. Rogers, M. Dymkov, K. Galkowski, D. H. Owens
42nd IEEE International Conference on Decision and Control (IEEE Cat. No.03CH37475), vol. 1, pp. 604–609 · DOI
2003
Robot optimal control in the presence of state constraints
S. Dymkou
Ballarat University Conference, Victoria, Australia
2002
Approximation of the optimal control problem with the nonsmooth state constraints
S. Dymkou
Industrial Optimization Symposium (WACEIO), Perth, Australia
2002
Suboptimality and differential properties of optimal solutions for repetitive control systems
S. Dymkou, L. Minchenko
Reports of the National Academy of Sciences of Belarus, vol. 46, no. 3, pp. 5–8 (In Russian)
2002
Optimal control for a class of differential linear repetitive processes
S. Dymkou, E. Rogers, M. Dymkov, K. Galkowski, D. H. Owens
15th International Symposium on Mathematical Theory of Networks and Systems (MTNS 2002), Indiana, USA (CD-ROM)
2002
Optimal control problem for multipass systems
S. Dymkou, L. Minchenko
International Conference “Erugin Readings VIII”, pp. 134–135, Brest, Belarus
2002
A linear optimization problem for repetitive processes
S. Dymkou
Bulletin of the National Academy of Sciences of Belarus, Physics and Mathematics Series, no. 4, pp. 35–41 (In Russian)
2001
Optimization problem for repetitive control systems
S. Dymkou, L. Minchenko
International Conference “Analytic Methods of Analysis and Differential Equations” (AMADE-2001), pp. 113–114, Minsk, Belarus (In Russian)
Education
Data Science Specialist (extended programme) — refreshing the toolset
2026 — present
Yandex Practicum · Online
PhD in Mathematics — Optimization and Optimal Control
2003 — 2006
RWTH Aachen University · Aachen, Germany
Thesis: Graph and 2D optimization theory for discrete simulation of gas transportation networks and industrial processes with repetitive operations. Grade 1 (highest, German scale).
MSc in Information Technology
2001 — 2003
University of Ballarat (now Federation University Australia) · Ballarat, Australia
MSc in Mathematics (engineer); additional qualification in Mathematics and Economics
1995 — 2000
Belarusian State University, Faculty of Mechanics and Mathematics · Minsk, Belarus
Certificates
GO IT! Data Science Embedded W2
Oct 2021 — Jan 2022
Raiffeisen Bank International (RBI), internal training
GoIT! Data Science Contribute Python W1
Sep 2020
Raiffeisen Bank International (RBI), internal training
Academic
Research interests
Optimization and optimal control; 2D and repetitive systems; control of networks (gas transportation); multi-agent systems and autonomous UAVs; machine learning, time series, NLP.
Honors & awards
2016 Seed Project Grant — Fast Multiple UAV Flight Generation Algorithms, Singapore
2003 DFG scholarship, RWTH Aachen, Germany
2001 Abroad scholarship, University of Ballarat, Australia
Talks
2014 Multifunctional transformation method in flow modelling. 19th IFAC World Congress, Cape Town, South Africa
2007 Optimal Control for Networks of Gas-Pipelines. ICIAM07, Zurich, Switzerland
2003 Approximations of the constrained nonsmooth optimal control problems and their applications for robot motion planning. ICIAM03, Sydney, Australia
Professional service
Reviewer: Unmanned Systems (2015–2017), Multidimensional Systems and Signal Processing (2007–2017), International Journal of Robust and Nonlinear Control (2007–2009), Systems and Control Letters (2007)
Member of Graduiertenkolleg RWTH Aachen (Germany, since 2003), Pacific Optimization Research Activity Group (Australia, since 2001), Belarusian Mathematical Society
