IDP with Parabella Analytics
📌 Key facts
uses the financial numbers to derive analytical insights and create a finance
report automatically with a “one click to report” feature. These automated
finance reports will be multi-validated in an AI-pipeline to become auditaligned
with the institutional frameworks.
Find all information in the PDF attached:
- 📌 Key facts
- 💡 Background
- 🎯 Goals
- 📝 How to Apply
- 📬 Contact
💡 Background
Transparent communication of annual financial numbers is essential for every
company, investor, regulator and stakeholders. The data input for a traditional
finance report is based on manually data collection and often not automated,
especially the textual interpretation of the numbers including forecasts, risk
assessments and so on. Parabella Analytics aims to automate this whole
reporting process by applying state-of-the-art AI techniques by extracting
necessary data automatically, represent analytical insights and converts all
financial data into coherent, narrative text of a finance report.
🎯 Goals
- Algorithm Development: Identification of a data-to-text algorithm like
BART or T5 to build a foundation framework that is specified on financial
reporting. Using training data to test efficiency and outcome-quality. - LLM Integration & Fine-Tuning: Evaluation of an AI-Architecture between
BART algorithm and GenAI models. Evaluate and text large-language
models to translate numeric inputs into coherent text at paragraph and
section levels. - Data Analytics & Schema Design: Define, how extracted, numeric data
can be used to build analytics frameworks and graphics. Identify a dynamic
report schema for multiple topics e.g. Management Discussion & Analysis,
KPI commentary, risk management, heatmap and so on. Creativity is
important. - End-to-End Automation: Design and implement a “one-click to report”
workflow that ingests raw financial statements and analytics-outputs in
form of a structured narrative report, without human intervention. You can
use the current AI-pipeline of the ESG-reporting module from Parabella
Analytics. - Data Validation & Consistency Checks: Incorporate rule-based and AIdriven
multi-validations with different GenAIs to ensure factual accuracy,
flag anomalies and maintain regulatory compliance for „audit-aligned
reports“
📝 How to Apply
If you are interested, please contact Prof. Dr. Isabell Welpe by submitting your CV and grade report. Please also briefly outline why you are interested and how you would approach the topic (data and methods, possible outcomes with a tentative outline all in word as *.docx)
We're greatly looking forward to hearing more about you!
📬 Contact
Isabell Welpe (Chair for Strategy and Organisation)
👉 welpe@tum.de