Carlos Henrique Tarjano Santos, PhD

• tarjano@live.com
• github.com/tesserato
• linkedin.com/in/carlostarjano
• orcid.org/0000-0003-3134-2778

LLMs, Multi-Agent Systems, MLOps, NLP, DSP

AI Systems Architect | ML Researcher (PhD) with experience in designing production-grade AI systems at the intersection of LLMs and DSP. Proven track record of translating advanced research into scalable infrastructure, including multi-agent LLM systems, feature stores, and synthetic data pipelines. Combines deep theoretical grounding with strong software engineering across Python, C++, and Rust, with a focus on building elegant, scalable and maintainable AI systems.

WORK

INESCTEC

Researcher (15/01/2024 - Present)

• Lead architect and developer of Voltus, a Feature Store for the CLOUDIA4.eu energy platform in the context of the ATE (Aliança para a Transição Energética) Portuguese project
• Designed an LLM-driven automated feature engineering system using multi-agent workflows and self-healing code generation in the context of the ATE project
• Led the implementation of synthetic time-series generation systems for AI-EFFECT (Horizon Europe TEF initiative), defining the architecture for the Portuguese node in coordination with partners including TU Delft (Netherlands) and Fraunhofer.
• Established MLOps practices and mentored junior researchers across reproducible pipelines
• Participated in conferences and workshops, presenting research and engaging with the scientific community

Match<IT>

Head of Data (17/03/2023 - 15/01/2024)

• Led the company’s AI/NLP strategy and delivery, overseeing both research and production systems
• Managed and mentored multidisciplinary teams of researchers and developers, aligning R&D with business objectives
• Coordinated funding activities, including the elaboration of successful research proposals and management of funded projects
• Oversaw talent acquisition, research roadmapping, and the transition of experimental NLP models into deployed products

AI/NLP Researcher (24/10/2022 - 17/03/2023)

• Designed and implemented NLP and similarity algorithms to match technology suppliers and buyers using structured and unstructured data
• Applied neural networks and modern NLP techniques to improve relevance, scalability, and robustness of matching pipelines
• Contributed directly to production systems, bridging applied research and real-world deployment

UFRJ (Federal University of Rio de Janeiro)

Academic Operations Officer (27/01/2009 - 24/10/2022)

• Held a tenured federal position obtained through a highly competitive national selection process
  
• Operated across multiple institutional roles within one of the largest universities in Latin America, combining administrative leadership, data analysis, and process automation
• Led and supported digital transformation initiatives, including workflow automation, document digitalization, and internal information systems
• Designed analytical tools to assess academic performance and scholarly production, supporting strategic and evidence-based decision-making

Key roles and responsibilities included:

• Administration of the International Relations undergraduate program
  Responsibilities included visual identity creation, academic performance analysis, and evaluation of scholarly production metrics
• End-to-end management of the Philosophy graduate program
  Responsibilities included maintaining the program website, reporting to national evaluation bodies, and analyzing scholarly output
• Service as vice-chief of the IFCS financial center
  Assisted in budget planning, execution, and oversight of institutional projects
• Supervision of document flow and compliance processes within IFCS
  Implemented automation and digitalization solutions to improve efficiency, traceability, and institutional governance

EDUCATION

PhD in Industrial Engineering and Management (Revalidation 320240067419) U.PORTO (Porto University) (07/03/2024)

PhD in Production Engineering UFF (Fluminense Federal University) (01/08/2018 - 28/07/2022)

Visiting Research Fellow at ICCMR (Interdisciplinary Centre for Computer Music Research) University of Plymouth (01/01/2020 - 31/03/2020)

MSc in Production Engineering UFF (Fluminense Federal University) (01/03/2017 - 31/07/2018)

Bachelor’s Degree in Production Engineering CEFET (Federal Center of Technological Education) (01/03/2010 - 01/12/2015)

PERSONAL PROJECTS

2026 - Tect. Declarative Language for Architecture-as-Code and System Simulation that identifies architectural flaws early in the design phase.

2025 - CodeWeaver. Command-line tool that transforms your codebase into a single, navigable Markdown document.

2025 - CloudMapper. Command-line utility designed to help you understand and Analyze your cloud storage.

2024 - Dead Simple Blog Generator. Command-line tool that transforms a directory of Markdown and/or HTML files into a static website.

2023 - Piano Forte. Piano synthesizer based on a lightweight (~600KB, ~150k parameters) neural network and a novel representation for Quasi-Periodic signals.

2022 - Omnes Sonos. The first plugin to leverage neural networks to sound synthesis in real-time.

2022 - Harmonic Compression. Lossy compression codec based on pseudo cycles segmentation. Registered in Brazil under the number BR512022000697-9.

2021 - Mark Book. Cross-platform songbook developed in Flutter. Available for Windows, Android and Web. Registered in Brazil under the numbers BR512022000618-9 BR512022000620-0.

2021 - Signal-envelope. A Python module to extract the envelope of digital signals. Registered in Brazil under the number BR512022000622-7.

2021 - BibRust. Bibliographic manager made in Rust. Registered in Brazil under the number BR512021000226-1.

2020 - Neural Piano. Neural networks based piano software. Registered in Brazil under the number BR512020001653-7.

2018 - SongB. A minimalist songbook app with chord detection, easy chord repositioning and optimized C++ code. Registered in Brazil under the number BR512020001654-5.

PUBLICATIONS

Valdecy Pereira | Marcio Basilio | Carlos Tarjano : Enhancing decision analysis with a large language model: pyDecision, a comprehensive library of MCDA methods in Python. Journal of Modelling in Management, August 2025. doi.org/10.1108/JM2-04-2024-0118.

Valdecy Pereira | Marcio Basilio | Carlos Tarjano : PyBibX - a Python library for bibliometric and scientometric analysis powered with artificial intelligence tools. Data Technologies and Applications, Volume 59, Issue 2, January 2025. doi.org/10.1108/DTA-08-2023-0461.

Carlos Tarjano : A general algorithm for the real-time emulation of pitched musical instruments and the singing voice. PhD Thesis - Production Engineering - Fluminense Federal University, 2022. dx.doi.org/10.22409/TPP.2022.d.11839251743.

Carlos Tarjano | Valdecy Pereira : An Efficient Algorithm For Segmenting Quasi-Periodic Digital Signals Into Pseudo Cycles: Application in Lossy Audio Compression. IEEE/ACM Transactions on Audio Speech and Language Processing, Volume 30, May 2022. doi.org/10.1109/TASLP.2022.3171969.

Carlos Tarjano | Valdecy Pereira : Envelope estimation using geometric properties of a discrete real signal. Digital Signal Processing, Volume 120, January 2022. doi.org/10.1016/j.dsp.2021.103229. Early manuscript available at arxiv.org/abs/2009.02860.

Carlos Tarjano | Valdecy Pereira : Signal-Envelope: A C++ library with Python bindings for temporal envelope estimation. arXiv, October 2021. doi.org/10.48550/arXiv.2110.11807.

Carlos Tarjano | Valdecy Pereira : Neuro-Spectral Audio Synthesis: Exploiting characteristics of the Discrete Fourier Transform in the real-time simulation of musical instruments using parallel Neural Networks. Lecture Notes in Computer Science Volume 11730, September 2019. doi.org/10.1007/978-3-030-30490-4_30.

Carlos Tarjano | Valdecy Pereira : Simulação de instrumentos musicais acústicos em tempo real utilizando redes neurais paralelas no domínio da frequência. Anais do XXXIX Encontro Nacional De Engenharia De Produção - ENEGEP, 2019. dx.doi.org/10.14488/ENEGEP2019_TN_STO_292_1654_37336.

Carlos Tarjano | Ruben Gutierrez | Valdecy Pereira : Traditional and alternative bibliometric indicators: assessing the evolution of interest in project management paradigms. Anais do XXXIX Encontro Nacional De Engenharia De Produção - ENEGEP, 2019. 2594-9713.

Carlos Tarjano : Redes neurais aplicadas à modelagem de instrumentos acústicos para síntese sonora em tempo real. Master’s Degree Thesis - Production Engineering - Fluminense Federal University, 2018. dx.doi.org/10.22409/TPP.2018.m.11839251743. Available online(Portuguese).

Carlos Tarjano | Valdecy Pereira : Neural Networks as a Tool for Big Data: The State of the Art. Anais do I Simpósio Internacional de Network Science - SINS, 2017. networkscience.com.br/anais/.

Carlos Tarjano | Magda Leite | Paula Purcidonio : O Processo de Desenvolvimento de Produtos em Startups Baseadas em Inovação Criativa. Espacios. Vol. 37 (Nº 07). Pág. 6, 2016. 0798 1015.

MISCELLANEOUS

2025 - Conference. Energy Technology Open Day. Presentation: Language Models for Automatic Feature Engineering. Porto, FEUP (October 15th).

2025 - Conference. Conferência ATE: Liderar a mudança: soluções para um novo paradigma energético. Stand showcasing the CLOUDIA4.eu platform MVP, the Voltus Feature Store, and synthetic time series generation results. I was particularly responsible for the Voltus Feature Store. Porto, Super Bock Arena (June 26th).

2025 - Blog. The Fortnightly Tambourine. A blog with the objective of popularizing science.

2025 - Peer Review. Article peer Review for EURASIP Journal on Audio, Speech and Music Processing.

2025 - Peer Review. Article peer Review for Journal of informetrics.

2023 - Competition. Neural Audio Plugin Competition. Organized by The Audio Programmer and Neutone. Presentation of the Piano Forte project.

2023 - Funding. Auxílio à Pesquisa - Pesquisa Inovativa em Pequenas Empresas - PIPE. Elaboration of a successful funding proposal for the Brazilian Agency FAPESP, securing a total funding of R$ 1.000.000,00 (Approximately 160.000 Euros). More details at orcid/funding.

2023 - Interview. Adeus aos longos formulários: como especificar projetos com conversas simples. Interview as Match<IT> Head of Data about conversational agents.

2019 - Conference. XXXIX Encontro Nacional de Engenharia de Produção (ENEGEP). Oral presentation of the paper “Simulação de instrumentos musicais acústicos em tempo real utilizando redes neurais paralelas no domínio da frequência”.

2019 - Conference. XXXIX Encontro Nacional de Engenharia de Produção (ENEGEP). Oral presentation of the paper “Traditional and alternative bibliometric indicators: assessing the evolution of interest in project management paradigms”.

2017 - Conference. I Simpósio Internacional de Network Science (SINS). Oral presentation of the paper “Neural Networks as a Tool for Big Data: The State of the Art”.

2014 - Crowd Funding campaign. Carcara Eco Footwear. Crowd Funding campaign for a line of eco-friendly shoes.

SKILLS