• Streamlit Expected Credit Loss dashboard with python calculations and SQL database connections.• Kubernetes deployment of Expected Credit Loss containing Greenplum and Oracle database connections.• Kubernetes deployment of macro-data dashboard containing proxy settings.

• Dashboard for macroeconomic data pulled form OECD and FRED.• Asset depreciation curves.• Impact of depreciation curves on ECL.

• Predicting daily demand for each product that Romstal has : http://freefullmotorsports.com/.• Creating webpages where they can download only those graphs which provide insight• Finding orders with abnormal number of items, profits, profit margins. Finding months in each warehouse with abnormal average number of orders, profitsand profit margins

• Working on a webpage which makes predictions for the number of severely infected people in every Romanian county, for example. It also makes other types of predictions and it creates graphs.• ML (KModes) for grouping employees• Finding employees that have high salaries or low salaries for their group.

• Gathering macroeconomic data.• Tried an AI that would figure out what to send to clients (SMS, letter, voice call, robot) that would maximize the chance that a client would pay and also maximize the amount of money collected from the client, while minimizing the cost.

• Login page using Google Authenticator’s QR code.• Wikipedia image and face parser.• Google’s and Facebook’s ViT transformers for facial recognition.• Github: https://github.com/0ff1c1ally-n0b0dy

• Wikipedia name parser.• Transformer neural network for translating names from English to Russian.

• Worked in Fusion a little bit.

• LSTM neural network for identifying personal identifiable information from json files with 98% accuracy.• LSTM neural network for identifying if a first name is an actual person’s name using a dataset with names from 17 countries with an accuracy of 92%.

• Long run average for probability of default.• Climate change stress test for individual economic sectors. Macroeconomic model taken from a British Research Institute which give projections for employment, GDP, etc. depending on carbon tax emissions that governments impose. Those macroeconomic predictions were digressed to individual sectors by using a model published by Banque de France.

• Created another alternative to the Black‑Litterman models I worked on during my Ph.D. This alternative incorporates a stochastic volatility model on the data’s common factors. In this way, the covariance matrix of the data is not only random, but also changes with time. Paper published in the InternationalJournal of Statistics and Probability: https://www.ccsenet.org/journal/index.php/ijsp/article/view/0/44571.• Taught Calculus 2, Linear Algebra 2, Regression Analysis and Applied Statistics I.

• A Bayesian Approach for Asset Allocation– Mathematically derived two variations of a model that takes as inputs both historical data and any prior information that the user might have about the linear relationships between the means of different populations. It outputs posterior means, posterior covariance and possible percentage weightsfor each population. The models maximize the posterior mean, minimize the posterior covariance and the Bayesian risk.– Complete work published in the International Journal of Statistics and Probability: http://www.ccsenet.org/journal/index.php/ijsp/article/view/0/42740• A Bayesian Approach for Large Asset Allocation.– Encountered running time and memory allocation problems when applying one of the Bayesian Alternatives from above to almost 500 variables. This is because a 106GB random matrix has to be generated at each iteration in a Gibbs Sampler.– Reduced the dimension by introducing Bayesian factor models to the already existing alternatives and implemented them.– Complete work published in the International Journal of Statistics and Probability: https://ccsenet.org/journal/index.php/ijsp/article/view/0/44345

• Customized tutorials for students to learn Python and SAS.• Facilitated discussions between students on the Piazza forum and in person.• Graded exams and assignments.

• Lead lectures for an introductory class offered by UCSB’s Statistics and Applied Probability Department‑PSTAT 5A.• Strategically customized content for 70 students accounting for varied proficiency levels and backgrounds.