Stavros Tsalidis

Employed as Data and Senior Data Scientist applying modern analytical approaches to solve industrial problems in diverse sectors. Recent example projects:

• Multi-objective optimisation for networks of health providers. Implemented and compared models: Integer Programming (IP), Genetic Algorithms, Simulated Annealing (also quantum version in @DWave). Networks optimised for cost efficiency and quality, accessibility. IP solution proved most performant and centrepiece of consuming web app.
• Call room scheduling optimisation; Implemented Erlang algorithms and @simpy simulations of call room operations, tuned and validated by real data. Demonstrated improvement in scheduling plans.
• Reinforcement Learning solution for scheduling of international liner itineraries to minimise shipping costs,
• Next action optimisation for marketing new drugs: implemented time to event prediction in R with dynamical survival models to estimate effect of marketing actions on adoption of drugs.
• Real time C++ implementation of particle filtering algorithms for positioning racing cars during Formula 1 races by data fusion from multiple sensors.
• Demand forecasting for several thousands of items of Walmart chains with DL models combining CNN and RNN layers. Ranked among top 4% of M5 competition with more than 5k teams participating.
• Inventory optimisation for aircraft service parts (Aftermarket) involving thousands of intermittent items as part of supply chain optimisation project: used appropriate version of @DeepAR model for demand forecasting which showed higher performance and replaced previous client solution.
• Racing strategy for electric car racing: simulated full race based on simulated car data and estimated optimal racing strategies using reinforcement learning (fitted Q-learning).
• Scheduling steal casting lines: simulated casting lines and optimised decisions (choice of alloy types and slab sizes) using reinforcement learning algorithms.
• RNA design: mRNA folding prediction and inverse folding. Compared recent approaches as Lineardesign, AntARNA, aRNAque etc, mostly implemented in C++.
• Fine-tuned protein language model ProGen2 to generate particular domain proteins using Hugging Face transformers with deepspeed and peft (LoRa) to parallelise and train on restricted GPU resources.
• Code development was carried out in collaborative setting using GIT and included high-quality model documentation for clear presentation to both technical and non-technical audiences.

• Risk management: built Real Time .Net application to follow betting patterns in games and provide continuous estimates of running state of betting (gains, losses, etc.) Identified possible scenarios of combinations of game outcomes that would result in important losses and issued warnings.
• Programmed private cryptographically strong Random Number Generator (RNG) in C++ that are used to sample winning numbers.
• Built tools in C++ for testing of randomness of samples generated by company's RNGs and obtained official certifications for use of these machines.
• Implemented statistical models (based on bivariate Poisson distribution) for estimating probabilities of scoring goals in soccer games used to evaluate betting payoffs.

Universities employed:

• 2003 - 2005: Brown University (Providence, Rhode Island)
• 2002 - 2003: University of Muenster (Muenster, Germany)
• 2000 - 2002: Aberdeen University (Aberdeen, Scotland)
• 1999 - 2000: University of Strasbourg (Strasbourg, France)
• 1997 - 1999: University of Oslo (Oslo, Norway)
• 1996 - 1997: Brown University (Providence, Rhode Island)
• 1994 - 1996: Purdue University (Lafayette, Indiana)

Academic work:

• Research in Mathematics and publications of results
• Presentations of work in seminars, colloquies, conferences.
• Taught undergraduate courses in Advanced Calculus, Linear Algebra, Probability and Statistics and graduate course on Algebraic Topology.

Publications:

• Tsalidis S, On the Descent Problem for Profinite Group Actions in Homotopy and Algebraic K-Theory, Aberdeen Topological Centre (ATC) No 10 (2002)
• Tsalidis S., On the Algebraic K-theory of Truncated Polynomial Rings, Aberdeen Topological Centre (ATC) No 9 (2001)
• Tsalidis S., On the Etale Descent Problem for Topological Cyclic Homology and Algebraic K-Theory, K-theory Vol. 21, No.2, October 2000, pages 151-199
• Tsalidis S.,Topological Hochschild Homology and the Homotopy Descent Problem, Topology Vol 37, No 4, pp. 913-934, (1998)
• Tsalidis S., On the Topological Cyclic Homology of the Integers, American Journal of Mathematics Vol. 119, pp. 103-125, (1997)