Marcelo Caro

About Marcelo Caro

Experienced Software Architect with a demonstrated history of working in the software industry. Skilled in Kubernetes, DevOps, Python, Haskell, Linux and Scrum methodology. Strong engineering professional with a BS focused in Computer Science from Universidad Nacional de Córdoba.

Experience

1. Architect Engineer Sr Principal Eng

   Proofpoint

   Jan 2024 — Present

   Seasoned engineering leader with extensive experience in Deep Learning, Large Language Models (LLMs), and scalable MLOps solutions.Proven track record of architecting and deploying AI models that improve detection accuracy and drive impactful results in enterprise environments.Expert in developing innovative NLP techniques like semantic compression and classification using LLM embeddings to extract meaningful insights from data.Strong technical leadership in building frameworks for research-to-production pipelines across multi-region, multi-cloud setups, leveraging tools like PyTorch, Amazon SageMaker, and Amazon Bedrock.Projects & AchievementsAI-Powered Threat Detection Improvements: Developed and integrated advanced deep learning models (using PyTorch) into Proofpoint’s email security platform. LLM-based Semantic Compression System: Created a novel semantic compression pipeline leveraging transformer-based LLMs (e.g, GPT-style models). The system compresses email text into dense vectors that retain contextual meaning, achieving substantial data compression.Global ML Research and Deployment Framework: Engineered a cross-cloud MLOps framework enabling one-click deployment of machine learning models across multiple AWS regions and cloud environments. This framework standardized model packaging, testing, and release processes, greatly accelerating deployment frequency and ensuring consistency and reliability of models in production.Semantic Embedding Classification Engine: Implemented a content classification engine using LLM-generated embeddings to group and label messages by topic and threat level. By leveraging semantic similarity of embeddings, the engine improved classification precision and recall, allowing security analysts to focus on high-priority threats with greater confidence.

Education

• UNC

  Bachelor of Science (BS), Computer Science

  2000 — 2006

• Universidad Nacional de Córdoba

  BS, Computer Science

  2000 — 2007

Skills

• Asm
• Subversion
• C
• Distributed Systems
• Object Oriented Design
• Multithreading
• Computer Security
• Unix
• Java
• Embedded Software
• Perl
• Software Design
• Haskell
• Bash
• Clearcase
• Shell Scripting
• C++
• Software Engineering
• Debugging
• Linux Kernel
• Linux
• Software Development
• Python
• Scrum