What is Gilion?

At Gilion, we’re building the world’s largest Growth Platform for founders. Our technology is at the heart of everything we do, enabling entrepreneurs to achieve their goals faster through innovative funding and analytics solutions. Our platform seamlessly connects raw business data to cutting-edge growth models, offering founders unparalleled insights and access to growth loans that are unmatched in the market.

We're a group of engineers hailing from leading innovators like Google, Spotify, King, Anyfin, and Einride. We've joined forces with some of the brightest minds in banking to assist entrepreneurs in growing their businesses smarter and faster. Traditionally, founders have been left to navigate the fundraising journey alone.

The Gilion platform offers immediate access to a powerful analytics tech stack, featuring intuitive visualizations and advanced forecasting without needing SQL or CSS. Our data-driven approach helps founders understand their performance compared to peers, as the industry shifts towards more objective, data-backed investment decisions. We’re fully committed to driving a new era for how entrepreneurs grow their ventures on a global scale.

Master Thesis

Gilion is a company dedicated to providing advanced forecasting solutions to help businesses make data-driven decisions. Accurate forecasting of customer acquisition is essential for making informed strategic choices. Therefore, customer acquisition forecasting forms the foundation of Gilion’s forecasting platform. As markets become increasingly complex and competitive, refining forecasting models is crucial for sustaining a competitive edge. In this thesis, we will provide an in-depth exploration of advanced time series forecasting techniques aimed at enhancing the accuracy and reliability of customer acquisition predictions. The focus will be on developing methods to detect trend changes, simulate new market conditions, and quantify the uncertainty in forecasts.

Research Questions

The master's thesis student will focus on addressing the following critical questions:

1. Reactiveness to Temporal Features: How can we rapidly and accurately react to cyclical patterns and trend changes triggered by significant events (e.g., market shifts, economic fluctuations)? What are the most effective methods to integrate these detections into our forecasting models to maintain or improve accuracy?

2. Limited Historical Data: What approaches can we use to effectively forecast customer acquisition in both existing markets with long enough historical data and also new markets with few or no historical data?

3. Exploring Lead Indicators: Metrics like marketing spend could play a key role in forecasting acquisition. How can we incorporate such signals in a flexible yet robust fashion?

4. Probabilistic Forecast: What is the confidence level of our customer acquisition forecasts? How can we quantify this uncertainty and improve the reliability of these predictions?
   

Methodology Overview

master's thesis student will have the autonomy to apply a range of traditional machine learning and deep learning techniques to address the research questions in realistic scenarios. Specifically, the student will:

1. Investigate state-of-the-art methods for trend change detection, including change point detection algorithms and anomaly detection models.

2. Implement a framework to train, evaluate, and compare these cutting-edge methodologies using real-world datasets.

3. Propose and experiment with new approaches aimed at delivering a robust and scalable forecasting solution.

4. Evaluate the effectiveness of the proposed solutions within Gilion’s existing framework.

Expected Outcomes

By the end of the thesis, the student is expected to deliver:

• A robust and scalable forecasting solution that significantly enhances the accuracy of customer acquisition predictions.

• Comprehensive documentation detailing the methodologies employed, the rationale behind their selection, and guidelines for future improvements.
  

Required Skills
ideal candidate for this master's thesis should possess the following skills:

• Proficiency in Python

• Experience with SQL

• Strong understanding of machine learning methodologies

• Knowledge of at least one deep neural network (DNN) framework, such as PyTorch or TensorFlow

• Familiarity with statistical forecasting tools

Sounds interesting? We look forward to your application!