Aaron Saeb

- Pre/post-training of LLMs at scale - Synthetic data generation and data mixing strategies - SWE agents

Design Deep Learning and ML models for Health care data. Applying NLP techniques and transfer learning to extract insight from unstructured call notes. Engaging with internal customers in the marketing and brand teams and helping them identify the key performance indicators (KPIs) • Build TensorFlow input pipeline for text data from set of unstructured call notes from REPs and HCPs touch-points using tf.data API. Tokenize and preprocess text for BERT ( Bidirectional Encoder Representations from Transformers) model. Fine tune BERT model with TensorFlow2 and TensorFlow Hub • Design and provide ML and pattern recognition techniques to gain marketing insight from Panel/Longitudinal data with both geographic and temporal components. • Work with research, business, brand, and software teams at Novartis and collaborate with academic institutions.

• Developed a sub-resolution model of binary black holes in the Illustris cosmological hydrodynamic simulation • Co-developed Python scripts to analyze the next generationWTNG simulation data (∼100TB) • Performed analysis of Illustris and TNG simulations data on large Linux clusters with 3000+ Nodes

•Taught 7 semesters of introductory physics lab sections with up to 60 students per semester •Prepared and troubleshot experiments, equipment and demonstrations. Initiated and moderated classroom discussions