about work contact

Marcos Esteve Casademunt

Senior Applied AI/ML Scientist I — Generative AI, Agentic AI, RAG, Search & Relevance, ML/NLP Engineering

Work

Senior Applied AI/ML Scientist I @ Thomson Reuters

Zug, Switzerland

Nov 2024 - Present

 Eval-Driven Development and agentic solutions
 Fine-tuned small language models for reranking and tool selection using synthetic data distillation from LLMs
 Autonomous agents for document modification
 Production multi-agent system with API tool-calling and internal/external search
 RAG pipeline for long-document retrieval and reasoning validated by subject-matter experts
 Document reranking for customer support with representation learning and cross-encoders
 Published and presented at EACL 2026

Machine Learning Engineer @ Whalar

Apr 2023 - Nov 2024

 LLMs
 Deep Learning (Tensorflow, Keras, PyTorch)
 Machine Learning (Scikit-Learn)
 NLP (Spacy, NLTK, Gensim)

ML Team Lead @ RavenPack

Jun 2021 - Apr 2023

 Deep Learning (Tensorflow, Keras, PyTorch)
 Machine Learning (Scikit-Learn)
 NLP (Spacy, NLTK, Gensim)

Machine Learning Engineer @ SolverIA

Sept 2020 - Jun 2021

 Deep Learning (Tensorflow, Keras)
 Machine Learning (Scikit-Learn)
 Time series
 Docker
 FastAPI

Data Scientist @ Datamaran

Sept 2019 - Sept 2020

 Machine Learning (Scikit-Learn)
 NLP (Spacy, NLTK, Gensim)
 MongoDB

About

Hi 👋🏻

I'm Marcos, an AI/ML Engineer with 8+ years of experience turning ML/NLP research into production systems — from Search & Relevance to Generative AI and Agentic AI.

I currently work on agentic systems and RAG pipelines for long-document understanding, fine-tuning small language models for tool-calling and reranking, and improving retrieval quality through representation learning. Some of this work is published at EACL 2026 and runs in production for subject-matter experts.

My background includes a degree in Computer Science and a Master's Degree in Artificial Intelligence, Pattern Recognition and Digital Image from the Polytechnic University of Valencia.

Currently my interests are: