2023-2024

AI Engineer at Incelligent

• Retrieval-Augmented Generation (Tech Consultancy): Led the research and development of a RAG Prototype, which included document ingestion, language-agnostic text processing, structural and semantic chunk segmentation, hybrid search, reranking, LLM Q\&A, benchmarking. • Retrieval-Augmented Generation (Multinational Bank): Developed an end-to-end RAG Chatbot, which combined novel AI and information retrieval methods to use relevant facts from a large corpus (10000 documents) as knowledge base to ground LLMs on accurate answers. • Document Information Extraction (Real Estate Marketplace): Contributed to the development of a document information extraction pipeline, which analyzed extensive property valuation reports and successfully extracted several valuation metrics.

2022-2023

Graduate Research Assistant at UT Austin NLP Lab

• LLM Optimization: Developed an effective method for LLMs (GPT‑4, GPT‑3.5) to detect and correct their own errors, by combining two differently prompted instances of the model in order to self‑reflect and emulate the human cognitive process associated with controlled reasoning. • Text Generation and NLI: Improved the performance of AI text generation models (GPT‑3, GPT‑J) by incorporating natural language inference in order to dynamically adjust the decoding strategy parameters and perform controlled text generation. • Data Augmentation and Contrastive Learning: Increased the effectiveness of AI natural language inference models (BERT, BART, ELECTRA) by proposing a novel method of data augmentation, combined with contrastive learning and a hybrid loss function.

2019-2021

Data Scientist at Pharmasept

• Recommendation System: Conceptualized and developed a recommendation algorithm, based on machine learning and time‑series forecasting, to automatically schedule the routes of sales consultants to their networks, thus minimizing risk and increasing annual revenue by 11%. • Budget Forecast: Designed the budget forecast for the sales department using predictive analytics, resulting in a sales target for each product. • Data Analysis: Utilized statistical techniques, such as a weighted ABC analysis and K-means clustering, for product and customer segmentation. • Stakeholder Advice: Communicated findings and visualizations to advise the stakeholders on strategic decisions for the company.