Hello, I'm William Beckhauser, a data science researcher and consultant helping product teams extract value from large volumes of data.
Generative AI | Machine Learning | NLP | Data Analysis | Knowledge Graph
hello@williambeckhauser.com
My published articles explore generative neural networks through knowledge graphs, reinforcement learning, and in-context learning. On the data-engineering side, I specialize in designing high-performance search systems and building data-centric platforms
Intelligent platform that helps researchers find the most suitable journals for publishing scientific articles, using AI and data analysis.
Framework that combines Retrieval Augmented Generation (RAG) with knowledge graphs and in-context learning to enhance the performance of language models.
Machine Learning, NLP
Ensuring the viability of large language models (LLMs) in situations requiring data privacy with limited on-premise resources is a significant current challenge. This work investigates how to tackle this challenge using knowledge graphs (KGs) and reinforcement learning (RL) to enhance minor LLMs. More.
Machine Learning, NLP
Traditional customer churn prediction uses profile and transaction data, leaving unexplored textual features such as customer reviews. This work compares machine learning models for churn prediction that use conventional data with those that use reviews posted by customers about their purchases. More.
Machine Learning, NLP
This study examines the application of Reinforcement Learning (RL) in conjunction with Large Language Models (LLMs) for the multi-label classification of financial news. The research evaluates models of different scales, including GPT-4 and Llama3, and constructs a dataset of 111,000 articles. More.
Florianópolis, Brazil
Bratislava, Slovakia
Florianópolis, Brazil
Brazil
The thesis focuses on improving smaller, open-source LLMs by using reinforcement
learning, knowledge graphs, and RAG. Preliminary results show a more
than 50% accuracy improvement in applications like information extraction.
Publications:
- EchoRAG: A Framework for Enhancing Language Models with Graph-RAG and
In-Context Learning
- OndePublicar.com: A Collaborative Web Platform for Dissemination and Curation
of Academic Events and Journals
- Innovation Award from the Brazilian Computer Society for OndePublicar
- Boosting not so Large Language Models by using Knowledge Graphs and
Reinforcement Learning
- Financial News Classification Using Language Learning Models and Reinforcement
Learning
Portugal, Spain
Led data science strategies for launching and migrating apps for major brands
(KFC, Pizza Hut, Ribs, Taco Bell), using unsupervised
models, segmentation, and recommendation systems. Utilized Python and SQL,
recovering 90% of active users and reactivating 30% of the
customer base, surpassing 650K users.
Created dashboards in Metabase, PowerBI, Looker Studio and Python (e.g my
dashboard for Brazilian researchers to find journals/conferences).
Conducted digital performance analysis for own and competitor brands using web
scraping. Analyzed and classified customer comments
with Llama and Knowledge Graphs to map issues across platforms like Google Maps
and Apple/Google Store.
Brazil
Dissertation focused on comparing conventional ML models and NLP techniques,
using customer textual reviews for churn prediction.
Publications:
- Churn Prediction in Enterprises with High Customer Turnover
- Can a Simple Customer Review Outperform a Feature Set for Predicting Churn?
Portugal, Spain
Coordination and development of models for menu creation on the website and
daily sales forecasting, reducing waste by over 45%.
Implementation of systems to predict churn, customer segmentation, predict
profitability, and product recommendations.
In-depth analysis of performance, revenue, marketing platforms, company
programs, and accounting.
Text analysis, customer categorization, and model building to predict churn and
systematically solve issues without human intervention
(Resulting in a scientific publication: "Can a simple review outperform a
feature set for predicting churn?").
