Security and governance
Prompt-injection defense, least-privilege tool access, per-customer data isolation, LGPD, ISO/IEC 27001 and 27701, NIST CSF, incident response.
Applied AI architect and CTO
From security practiceto AI in production.
I’m Hugo Frauches. I design and run agentic AI for security operations, from the executive conversation to production, with the privacy, evaluation and human controls that regulated teams expect.
Based in Brazil · UTC−3
Programs and rooms
- Conecta TI 2026 · Natercoop
- Bandes
- TIES / Data Saturday 2026
- Amcham Espírito Santo
- BSides Vitória 2026
- ISA Cyber Official Launch
- I Workshop de Tecnologias de Redes do PoP-ES
01 / What I do
What I work on.
Agentic systems, retrieval, evaluation and guardrails.
Agentic AI
OpenAI API, tool calling, structured outputs, retrieval-augmented generation, evaluations, LangChain and LangGraph, guardrails.
Architecture
Python, APIs, Kubernetes, Azure, AWS and GCP, CI/CD, and monitoring of AI services.
People in the loop
Privacy, least privilege, and a person on high-impact actions.
02 / Selected work
Three systems in production.
Case studies of AI products I conceived, helped build and led. No client names, and no private architecture.
ISA Horizon
AI external attack surface intelligence
An agent that maps external exposure and leaked credentials the way an attacker would look, with a person validating every finding before it is shared.
Read the case study (ISA Horizon)ISA Insight
AI vulnerability operations copilot
Turns findings from multiple scanners into risk-prioritized remediation plans, and answers questions about them in natural language.
Read the case study (ISA Insight)ISA Cyber
Agentic AI SOC platform
A multi-agent platform that triages, enriches, investigates and correlates security alerts, and opens qualified incidents around the clock.
Read the case study (ISA Cyber)Product names belong to their owner. No client data, and no private architecture, is published here.
03 / Talks & panels
Saying it out loud, in rooms that push back.
Conferences, workshops and executive rooms, in Brazil. Titles in English are editorial translations.
Security in the age of AI: protecting data from intelligent threats and leaks
Conecta TI 2026 · Natercoop
Deep and dark web: when data becomes currency
Bandes
Safe use of tools and protection of sensitive information
Bandes
Automating SAST & DAST with AI: how far can we go without a human in the loop?
TIES / Data Saturday 2026 · UniSales
Protecting business in the era of autonomous AI
Amcham Espírito Santo
The new era of the SOC: surviving AI-powered attacks
BSides Vitória 2026 · FAESA · Vitória, Brazil
Presenting ISA Cyber at its official launch
ISA Cyber Official Launch · Base27 · Vitória, Brazil
Strategies for the new attack surface
I Workshop de Tecnologias de Redes do PoP-ES · UFES · Vitória, Brazil
04 / The notebook
Thinking in public. Carefully.
Notes on AI, cybersecurity, product, and leadership. Primary sources, practical questions, and room to change my mind.
Security in the age of AI: protecting data from intelligent threats and leaks
The content of my talk at Natercoop’s Conecta TI 2026: what changed, where companies are exposed, and a 90-day plan to govern AI use.
AI · Security
Read the essay (Security in the age of AI: protecting data from intelligent threats and leaks)Leading also means making the decision nobody wants to make
Putting a decision off has consequences too. On judgment, firmness with respect, and staying accountable for what follows a hard choice.
People & leadership
Read the essay (Leading also means making the decision nobody wants to make)Published in English and Portuguese, after personal review.Subscribe via RSS →
05 / Recognition
A public record.
Certifications and study that sit alongside the work.
Check Point DemoPoint Rising Star, 2022.
Certifications
- CompTIA CySA+ (2023)
- CompTIA Security+ (2017)
- Check Point Certified Security Expert, CCSE R80 (2018)
- F5 Certified Administrator, BIG-IP (2018)
- Rapid7 InsightIDR Certified Specialist (2020)
- Fortinet NSE 3 (2026)
- Anthropic, Claude Code in Action (2026)
- Fortinet Certified Associate and Fundamentals Cybersecurity
- Microsoft Certified: Azure Fundamentals
Education
- Applied AI Engineering, postgraduateIn progress · Coders University · from October 2026
- Cyber Threat Intelligence, postgraduateInstituto DARYUS · 2022
- Game Development and Design, technology degreeFAESA · 2012–2014
06 / Conversations & connections
What could we build together?
A useful conversation starts with the problem, the constraint, and what a good outcome would change. For AI, security, product or a partnership, LinkedIn is the place to begin.
Thoughtful questions and shared experience at the intersection of AI, security and product.
Start a conversation (LinkedIn)A meaningful security challenge, a product idea, or something we could build together.
Explore a collaboration (LinkedIn)Complementary perspectives and capabilities with a clear shared purpose.
Start a partnership conversation (LinkedIn)LinkedIn is the best place to start. Mention what brought you here and which conversation you have in mind.