Miguel de Vega is a computer scientist and cryptographer specializing in privacy-enhancing technologies (PETs) and distributed systems. He is the co-founder and Chief Scientist at Nillion, where he leads research into cryptographic methods for secure, decentralized computation. ^[1]

Education

De Vega completed his Master of Engineering in Telecommunications Engineering at the Universidad Politécnica de Madrid (UPM) in 1999. He later earned a Doctor of Philosophy (Ph.D.) in the same field from the Université Libre de Bruxelles (ULB) in 2008. ^[2]

Career

De Vega began his career in academic and telecommunications research, contributing to European projects such as NOBEL and APSON, which were centered on developing next-generation optical networks. His work during this period involved the development and patenting of network architectures, with a focus on traffic engineering, optical burst switching, and quality of service (QoS) protocols for mobile and IP-based systems. His responsibilities included modeling, simulation, and performance analysis of network technologies.

Following his research career, de Vega transitioned into project management at Open Canarias, where he oversaw software development and modernization projects for financial institutions. He subsequently entered an entrepreneurial phase, co-founding several startups focused on web technologies and user interaction tools, including Browseye, iFlikeU, and Dialective. His career then shifted toward data science when he became the Vice President of Data Science at Treexor. In this role, he led a team that applied statistical modeling, deep learning, and machine learning to fraud detection and business performance optimization.

His focus later narrowed to artificial intelligence and privacy-enhancing technologies. He served as an advisor for Botdreams, assisting in the creation of conversational AI tools for the hospitality industry. At the digital identity company Sedicii, he first served as a technical advisor before becoming its Chief Technology Officer (CTO). As CTO, he was responsible for developing solutions that used zero-knowledge proofs (ZKPs) and secure multi-party computation (MPC) for identity verification and compliance. In 2021, de Vega co-founded Nillion and assumed the role of Chief Scientific Officer. At Nillion, he leads the scientific efforts to build a decentralized infrastructure for secure data storage and computation by integrating various PETs. ^[3]

Interviews

Nillion

In a Mach 2025 interview with Proof of Coverage Media, de Vega discussed Nillion's progress in leveraging privacy-preserving technologies for AI applications. He highlighted the rapid deployment of the R1 model from DeepMind on their platform, enabling users to interact with AI in a privacy-conscious manner. The conversation also touched on the implications of open-sourcing AI models, particularly concerning user privacy. De Vega emphasized that integrating privacy features should be seamless for developers and users alike, suggesting that technologies such as Zero-Knowledge TLS could bridge the gap between Web2 and Web3 infrastructures while maintaining data protection. Furthermore, he noted the importance of communication between AI agents and the need for secure, privacy-preserving frameworks to enhance efficacy in agent interactions. The discussion concluded with de Vega expressing enthusiasm about the future of AI, specifically the need for privacy as AI technology becomes more pervasive in everyday life. ^[7]

Blind Warriors and Crypto

On the Block By Block Show in January 2025, de Vega detailed his transition from academic engineering and mathematics into the Web3 landscape during a podcast discussion. After working with major telecom companies such as Nokia and Siemens, he became fascinated by distributed networks. Eventually, he shifted to privacy-enhancing technologies after discovering zero-knowledge proofs in 2013. Nillion was founded in 2021 to create a decentralized infrastructure for handling private data using a unique set of privacy technologies distinct from traditional blockchain solutions. The conversation emphasized how Nillion is positioning itself in the privacy and AI niche and aims to attract developers by making complex privacy-enhancing technologies more accessible through modules and SDKs. The challenges of communicating Nillion's value proposition, particularly within the privacy sector, were also addressed, emphasizing the need to demonstrate practical use cases to help both Web2 and Web3 developers understand the potential of integrating privacy features into their applications. ^[8]

Future of Decentralization

In an LP Roundtable discussion on MIDCRUVED in January 2025, de Vega shared insights on decentralization, AI, and privacy. He recounted his journey from studying telecommunications and mathematics to becoming intrigued by distributed systems and cryptography, which laid the groundwork for the development of decentralized infrastructure. De Vega noted the evolution of AI from basic chatbots to more autonomous agents, likening their growth to a military hierarchy where general commands lead to decentralized execution. He emphasized the critical need for privacy in AI, especially as systems gather sensitive personal data. De Vega also highlighted the potential of decentralized technologies to enhance user experience and promote accountability, particularly in telecommunications and AI. As Nillion prepared for its imminent launch, he expressed excitement about the future applications of privacy-enhancing technologies and the role they will play in shaping decentralized AI solutions. ^[9]

Presentations

Enhancing TFHE Bootstrap

In November 2024, de Vega presented at the FHE Summit II about enhancing TFHE bootstrapping through two research papers, Ripple and Curl. These papers addressed the challenges of running AI using privacy-enhancing technologies, specifically focusing on evaluating nonlinear functions within fully homomorphic encryption (FHE) and secure multi-party computation (MPC) frameworks. He introduced the discrete wavelet transform (DWT) as a technique for compressing lookup tables used to evaluate complex functions, thereby improving efficiency without sacrificing accuracy. This approach helped overcome the scalability issues associated with traditional lookup tables, which tend to grow exponentially with the number of input bits. The research demonstrated significant advancements in both compression and speed, providing a framework for more efficient resource use in privacy-based AI applications, in collaboration with various academic and industry partners. ^[5]

Trust-Minimized MPC

De Vega presented on the development of trust-minimized multiparty computation (MPC) protocols at DeCompute in October 2023, discussing the integration of building blocks such as linear secret sharing, fully homomorphic encryption (FHE), and garbled circuits to create scalable, decentralized solutions. He reflected on the historical context of web development, emphasizing the shift from the public nature of the early internet to the current centralized Web2, and the challenges posed by the introduction of decentralized Web3 technologies, which often compromise confidentiality. De Vega outlined a framework for understanding decentralization across architectural, political, and logical dimensions, analyzing the implications of each in building efficient MPC networks while addressing challenges such as the civil attack and silent leakage of shares. He explored the advantages and limitations of each cryptographic primitive, suggesting that a combination of these technologies could offer the most effective solutions for diverse use cases in privacy-enhancing technologies. ^[4]

Panels

Personal Agents

At the Open AGI Summit in November 2024, a panel discussion on how personal agents might transform daily life was held. De Vega, Regan Peng (PinAI),  Alex Hicks (Ethereum Foundation), and Will Villanueva (BonkBot) explored the implications of AI-driven agents on privacy, reliability, and user experience. De Vega emphasized the need for confidentiality in personal AI, expressing concerns about how these agents could collect extensive personal information, particularly if operated on centralized infrastructures. Peng introduced the concept of a decentralized coordination layer for personal AI that would facilitate data collection across various sources while preserving user data sovereignty. Hicks highlighted the potential challenges of managing chains of AI agents completing tasks, as the reliability of the output could diminish with each additional agent. The discussion underscored the importance of balancing user experience and data privacy, acknowledging the risks of centralized control over personal information in the evolving landscape of AI technologies. ^[6]