Ruan Chaves Rodrigues

Ruan Chaves Rodrigues

Senior AI EngineerGenerative AIAgentsRAGLLMsNLPPython Brazil

Contact:

Summary

Senior AI Engineer with extensive experience architecting and deploying production-ready Generative AI, Multimodal AI, and NLP solutions for US-based tech companies. I specialize in building complex Agentic Systems and RAG architectures, bridging the gap between research concepts and scalable, enterprise-grade applications. I hold a Double Master’s degree in AI & NLP (University of Malta & University of the Basque Country) and a Bachelor’s in Computer Science. My work focuses on remote-first collaboration with distributed teams across the US, Europe, and LATAM, delivering high-impact AI/ML pipelines on schedule.

Technical Ecosystem:

  • Agentic AI & Orchestration: Expert in designing Multi-Agent Frameworks and Autonomous Agents using LangGraph, LangChain, and LlamaIndex. Proficient in low-code workflow automation with n8n.

  • Generative AI & LLMs: Deep expertise in Fine-tuning, Function/Tool Use, and Prompt Engineering with OpenAI, Anthropic, and AWS Bedrock APIs. Advanced knowledge of Graph RAG, Multimodal AI, Transformers, and standard GenAI libraries (Hugging Face, spaCy, Haystack).

  • MLOps & Engineering: End-to-end deployment using Docker, Kubernetes, Kubeflow, and MLflow for model lifecycle management. Strong background in A/B Testing and Model Evaluation (e.g., LangSmith).

  • Cloud & Data Infrastructure: AWS (SageMaker, Bedrock), GCP (Vertex AI, BigQuery), and Azure. Skilled in ETL processes and managing high-scale data with Apache Spark, SQL, and Vector Databases (Pinecone, Weaviate).

  • Core Stack: Python, PyTorch, TensorFlow, Keras, Scikit-learn, FastAPI, Node.js, React.

Experience

Xseed Solutions Senior AI Engineer September 2025 - Present (5 months) | United States

Qive Senior AI Engineer February 2025 - August 2025 (7 months) | Brazil

  • Led the development of a product search engine over proprietary data, reporting directly to the company’s founders in a fast-paced startup environment.

  • Reduced RAG indexing time from several hours to minutes without compromising retrieval quality by uncovering reliable product groupings through data analytics and targeted indexing strategies.

  • Increased RAG retrieval efficiency by achieving over 90% recall with only 20 candidates (down from 100), through a redesigned vector retrieval system using GCP tools and Gemini-based data augmentation.

  • Conducted technical interviews for software engineering candidates, assessing coding skills, system design, and cultural fit.

  • Mentored team developers to improve code quality, strengthen technical expertise, and support professional growth.

C6 Bank Total Duration: 1 year 5 months

  • Senior AI Engineer (May 2024 - February 2025)

  • Led AI initiatives within a cross-functional team at one of Brazil’s largest digital banks, serving over 30 million clients.

  • Collaborated closely with backend, MLOps, data science, sales, and marketing teams to develop scalable, high-impact AI solutions that drove business growth and enhanced customer experience.

  • Built and deployed a Generative AI sales assistant using a multi-agent architecture with LangChain and LangGraph. The solution reduced both customer acquisition costs by a factor of six, improving sales team productivity.

  • Led the full development lifecycle of AI agent systems, including stakeholder alignment, design, testing, and production rollout.

  • Applied a range of Agentic AI techniques such as tool use, custom workflows, secure execution guardrails, few-shot learning, personas, model selection, and REACT strategies to enhance system performance and reliability.

  • Data Scientist (October 2023 - May 2024)

  • Optimized a retrieval-augmented generation (RAG) system on Google Cloud Platform using advanced prompt engineering techniques and research into new GCP features, reducing inference costs by 90% and improving response accuracy beyond top competing LLM alternatives.

  • Enhanced customer segmentation insights and enabled targeted marketing campaigns by developing time series forecasting and spending profile clustering models using BigQuery ML, resulting in more precise customer profiles.

Argilla Data Science Intern November 2021 - May 2022 (7 months) | Spain

  • Worked closely with early members in a fast-paced startup (later acquired by Hugging Face).

  • Pioneered the first production-ready implementation of state-of-the-art embedding-based annotation methods in the Argilla open-source library, contributing 68K+ lines of code.

  • Translated cutting-edge academic research into scalable, robust features deployed in real-world NLP workflows.

  • Authored tutorials and documentation that boosted user adoption.

Centro de Excelência em Inteligência Artificial (CEXIA) & Deep Learning Brasil AI Engineer September 2020 - July 2021 (11 months) | Brazil

  • Collaborated in a research team bridging academia and industry at CExIA, contributing to early-stage Transformer and language model research during the initial release of these technologies.

  • Achieved 1st place twice in major NLP competitions open to both academic and industry participants:

  • ASSIN 2 (2019): Led a stacking ensemble of multiple Transformer models for textual entailment, surpassing the runner-up by 0.7% in F1-score. Competition organized by the Brazilian Computer Society (SBC).

  • ABSAPT (2022): Secured top accuracy in Aspect-Based Sentiment Analysis in Portuguese through meticulous hyperparameter tuning of Transformer models, beating the second-place team by 3.5%. Organized by the Spanish Society for Natural Language Processing (SEPLN).

  • Delivered proof-of-concept NLP solutions to Brazilian enterprises including Copel (Forbes Global 2000), applying cutting-edge techniques in intent detection, named-entity recognition, sentiment analysis, entity linking, and early generative chatbot models.

NeuralMind Data Scientist February 2021 - May 2021 (4 months) | Brazil

  • Won 1st place out of 7 teams in the COLIEE 2021 textual entailment competition by applying zero-shot Transformer models without domain-specific adaptations, surpassing the runner-up by 6% in F1 score.

  • The Competition on Legal Information Extraction/Entailment (COLIEE) 2021 was organized by the Alberta Machine Intelligence Institute (Amii) at the University of Alberta.

Education

University of Malta Master of Science in Human Language Science and Technology, Artificial Intelligence August 2021 - September 2023

Universidad del País Vasco/Euskal Herriko Unibertsitatea Master in Language and Communication Technologies, Artificial Intelligence August 2021 - September 2023

Universidade Federal de Goiás Bachelor’s degree, Computer Science January 2017 - June 2021

Skills, Honors & Publications

Top Skills

  • Natural Language Processing (NLP)

  • Machine Learning

  • Python

Languages

  • English (Full Professional)

  • Portuguese (Native or Bilingual)

Certifications

  • From Data to Insights with Google Cloud

  • NVIDIA DLI Certificate - Fundamentals of Accelerated Computing with CUDA C/C++

  • Modernizing Data Lakes and Data Warehouses with Google Cloud

Honors & Awards

  • 1st place at the II Evaluation of Semantic Textual Similarity and Textual Inference in Portuguese (University Council’s Certificate of Honours)
    • ASSIN 2 shared task at STIL 2019. Competed against 8 teams on ~10K annotated Portuguese sentence pairs for semantic similarity and textual entailment.
  • 1st place at the Competition on Legal Information Extraction/Entailment (COLIEE) 2021
    • Task 2 (legal case entailment). Organized by the University of Alberta. Zero-shot model outperformed fine-tuned DeBERTa/monoT5 by 6+ points on Federal Court of Canada case law.
  • 1st place at the Aspect-Based Sentiment Analysis in Portuguese (ABSAPT @ IberLEF 2022)
    • Organized by the Spanish Society for NLP (SEPLN). Achieved 0.82 balanced accuracy on hotel review sentiment using RoBERTa/mDeBERTa ensembles, beating 4 other teams.

Publications

  • Construção de Datasets para Segmentação Automática de Hashtags
    • EnAComp 2020. Proposes a heuristic method for building hashtag segmentation datasets from tweets, statistically improving deep learning models.
  • Yes, BM25 is a Strong Baseline for Legal Case Retrieval
    • COLIEE 2021 workshop. Demonstrates that BM25 remains competitive against neural retrievers in low-resource legal domains.
  • Multilingual Transformer Ensembles for Portuguese Natural Language Tasks
    • ASSIN 2 workshop (2020). Ensemble of translated English and Portuguese Transformers achieved best RTE results, outperforming BERT-multilingual without task-specific preprocessing.
  • Domain Adaptation of Transformers for English Word Segmentation
    • BRACIS 2020. Applies continued pre-training and vocabulary expansion to BERT for compound word segmentation.
  • Deep Learning Brasil at ABSAPT 2022: Portuguese Transformer Ensemble Approaches
    • IberLEF 2022. State-of-the-art results on aspect term extraction (RoBERTa + mDeBERTa) and sentiment orientation (PTT5 voting ensemble).