During my engagement with Toptal, I have collaborated on two projects. The first involved building a data cleaning pipeline for a TripAdvisor-style product, overcoming challenges like data compatibility, error correction, and duplicate detection. The second project focused on merging databases for a telecommunications company, utilizing SQL queries, geospatial data, and visualization tools to generate reports for the database merging process.Main Activities:TripAdvisor-Style Product:- Developed a data cleaning pipeline to extract unique business data from multiple databases, ensuring compatibility across different economies.- Implemented automatic processes to correct falsely input data and minimise reliance on Google API calls.- Detected and appropriately handled duplicates and chains in the merged dataset.- Built a BERT model in PyTorch to extract sentiment from user-generated reviews.- Built a model to extract business- and sector-specific keywords from user-generated reviews.Telecommunications Database Merging:- Wrote SQL queries in BigQuery and Amazon Athena to retrieve comparable data from databases of merging telecom companies.- Worked with geospatial data, including latitude, longitude, and geohashes, to analyze and visualize the data using Seaborn in Jupyter Notebook.- Generated reports outlining the recommended merging strategies for the two databases.Main Outcome:Successfully contributed to the development of a data cleaning pipeline and sentiment analysis model for the TripAdvisor-style product. Additionally, provided valuable insights and recommendations for merging databases in the telecommunications industry, facilitating a seamless integration process and enhancing data analysis capabilities.

As a part-time scientific researcher at HEIG-VD, I am actively involved in research on Model-X Knockoffs, a framework proposed by Emmanuel Candès in 2018, aimed at reducing the False Discovery Rate (FDR) in machine learning models. With the support of the Management and Engineering School of Vaud, my current focus is on applying this framework to index-replicating stock portfolios, particularly those utilising variational autoencoders. Through this supplementary role, I contribute to advancements in data science and financial modeling.Main Activities:- Conducting research on Model-X Knockoffs, a framework designed to mitigate the False Discovery Rate (FDR) in machine learning models.- Collaborating with the Management and Engineering School of Vaud (HEIG-VD) to support and contribute to the research project.- Applying the Model-X Knockoffs framework to index-replicating stock portfolios, with a specific emphasis on utilising variational autoencoders.- Exploring the effectiveness of this framework in enhancing the performance and reliability of machine learning models in the financial domain.Main Outcome:Through my voluntary role as a scientific researcher, I actively contribute to research efforts focused on Model-X Knockoffs. By applying this framework to index-replicating stock portfolios, particularly those utilising variational autoencoders, I aim to advance the field of data science and financial modeling. The outcomes of this research can potentially lead to improved models with reduced False Discovery Rate, providing valuable insights and practical applications in the domain of machine learning and finance.

Using data from the Swiss Market Index and its stocks' Option data, we built models that optimised the weights for a portfolio of stocks and options that maximised expected return on investment.

Leveraged machine learning and data science in iKentoo's point-of-sale software for the hospitality sector to enhance operational efficiency, automate tasks, and offer personalized services from 2018 to 2021.Main Activities:- Developed an app estimating waiting times in restaurants during peak hours using simulations based on conditional log-normal distributions, taking into account various factors such as table occupancy and capacity.- Trained a Natural Language Processing model to autocategorize support tickets, aiding the support staff in handling customer complaints and queries more efficiently across English, French, German, and Italian.- Built a cluster model for client preferences to offer bespoke services tailored to clients' specific subscription preferences.- Designed a revenue prediction model incorporating weather forecast data to predict future revenue for a diverse array of restaurants, adjusting for operational hours.Main Outcome:Significantly improved the efficiency and accuracy of restaurant operations and customer support while providing tailored services and anticipating revenue trends despite challenges posed by COVID-19 lockdowns.

During my tenure as the Head of End-User Engagement at Mandat International, I led the delivery of various reports for the EU's Horizon 2020 program. My responsibilities encompassed providing publicly accessible educational resources on topics such as IoT technology, Smart City interfaces, and the GDPR privacy legislation. Additionally, I delivered a comprehensive report on stakeholders within the global cybersecurity ecosystem. Through my role, I contributed to enhancing public awareness and understanding of these critical areas.Main Activities:- Leading the delivery of reports for the EU's Horizon 2020 program, focusing on end-user engagement.- Creating publicly available educational resources to promote citizen awareness and knowledge in areas including IoT technology, Smart City interfaces, and the GDPR privacy legislation.- Conducting research and analysis to compile a comprehensive report on stakeholders within the global cybersecurity ecosystem.- Collaborating with teams and stakeholders to ensure the timely and accurate delivery of reports as per program requirements.Main Outcome:As the Head of End-User Engagement, my work contributed to the dissemination of valuable educational resources and reports under the EU's Horizon 2020 program. By providing accessible information on IoT technology, Smart City interfaces, and the GDPR privacy legislation, I helped enhance public understanding of these topics. Furthermore, the comprehensive report on stakeholders in the global cybersecurity ecosystem facilitated a deeper understanding of the key actors involved in ensuring cybersecurity measures. Overall, my efforts aimed to improve end-user engagement and foster informed decision-making in the areas of technology, privacy, and cybersecurity.

Research Projects:1 Stock Market Irrationality Sentiment:• Built a time series regression model combining sentiment analysis of irrationality in financial press articles with stock returns from NYSE and NASDAQ.• Developed a sentiment measure by compiling a lexicon of irrationality-related words and analysing their presence in articles referring to the stock market.• Conducted Granger causality tests and applied Newey-West standard errors to analyse the relationship between sentiment and stock returns.2 Aesthetic Preferences in Music and Money-Sharing Experiments:• Created a VBA app for participants to rank musical intervals based on aesthetic appeal.• Utilised logistic regression models to investigate whether aesthetic preferences in music influenced selections in money-sharing experiments.• Found preliminary evidence that musically trained individuals may be more likely to select aesthetically-pleasing splits, suggesting the need for further research.3 Theoretical model for the Preference Reversal phenomenon:• Developed a theoretical model using the stochastic Stronger Utility preference model to address the Preference Reversal phenomenon in economics.• Modified the utility model to incorporate the endowment effect, which influences individuals' valuation of possessions.• Provided evidence that the endowment effect may contribute to the presence of the Preference Reversal phenomenon. Teaching:• Assisted in teaching Asset Pricing theory to Masters students studying Management.• Prepared and graded exercises and exams, and provided coaching for problem-solving.Main Outcome:Contributed valuable research insights in the areas of stock market sentiment, aesthetic preferences, and preference reversal in economics. Published papers in reputable platforms and received recognition through awards. Additionally, provided guidance and support to students as a course assistant, facilitating their understanding of Asset Pricing theory.

For a brief period, I provided interviewing and TV presenting services to Dukascopy Bank in Geneva. This role required me to research my interviewee, write questions, and conduct the interview. Interviewees included business-men and -women, artists, singers and dancers, and politicians, such as the CEO of the Vote Leave campaign during the run-up to the Brexit vote.

I studied topics on finance. This formed part of my preparation for the PhD in Geneva.

At Renault Finance I built a piece of software in Excel that read data from a spreadsheet provided by a risk manager, took inputs on order sizes in foreign currencies, and provided a comprehensive breakdown of the entire order's risk profile with a high level of customisation.

I studied the fields of popular psychology and economics and proposed scenarios for the pilot of an educational programme for The Discovery Channel. I also searched for and suggested potential hosts and presenters for the pilot.

During my university studies, I worked part-time at a shop selling videogames and accessories. I had to be informed about new releases, console specifications, and rewards systems while being simultaneously friendly, approachable, and diligent.