About the Architect
Gustavo Peixoto de Azevedo
Hands-on AI Solution Architect — production-grade, event-driven agentic systems, from business concept to production and evolution.
I design and build production agentic systems. Buyer Team is the most recent and my flagship — an autonomous multi-agent procurement-negotiation SaaS platform I architected and shipped end-to-end on Amazon Bedrock AgentCore and the Strands Agents SDK.
25+ years architecting distributed systems. M.Sc. Computer & Systems Engineering, UFRJ. Based in Blumenau, Brazil; working with teams worldwide.
The Buyer Team build
I designed and built every layer of Buyer Team — business architecture, agent orchestration, multi-tenant runtime, integrations, observability, evals, security posture, deployment, and cost engineering — it demonstrates what a senior architect can deliver from concept to production, and evolution.
The platform ingests purchase requisitions, classifies them on the Kraljic matrix (profit impact × supply risk), routes each one to the right negotiation strategy — spot bid, competitive auction, risk-managed partnership, or strategic engagement — and runs the full cycle end-to-end: supplier invitations, multi-round bidding, evaluation against governance rules, and Purchase Order assembly. All autonomous. All audited. Minimum AI cost per negotiation, achieved through a four-layer cost-optimization architecture on Amazon Bedrock.
Stack: Strands Agents SDK · Amazon Bedrock AgentCore (Runtime, Memory, Gateway, Identity) · Python (FastAPI) · Next.js · MCP · multi-tenant by design · OWASP Agentic 2026 security mapping · MITRE ATLAS threat model · 7-year immutable audit trail · GitOps with Terraform.
Selected experience
AI Solution Architect · Buyer Team
Designed and built Buyer Team end-to-end on Amazon Bedrock AgentCore and the Strands Agents SDK — business architecture, agent orchestration, multi-tenant runtime, MCP integrations (SAP / Oracle), evals and observability, security and audit posture, and < optimized negotiation cost engineering. See above.
AI Native Software Engineer · Independent
Conception, design, development, and evolution of production agentic systems — from scratch with native orchestration or on frameworks (AWS Strands, LangGraph) — deployed on AWS, Vercel, and on-premises. Python (FastAPI) and Java (Spring AI, Embabel) backends, Next.js frontends, MCP, RAG; shipped via GitHub Actions to AWS Lambda, App Runner, AgentCore Runtime, and Kubernetes, with observability, evals, GitOps, and Terraform IaC. Active open-source contributor (see below).
Software Engineer (Platform) · Grepr
Founding-team contract at a San Francisco observability-cost startup. Designed and developed a load-test tool for Flink jobs; built a parallel data-backfill solution from Iceberg tables to Flink data pipelines; designed a new Flink connector.
Senior Software Engineer · Independent
Architectural reviews for early-stage startups. Built event-driven microservices systems on Spring Boot / Quarkus deployed to AWS. Worked extensively with OpenTelemetry, GraphQL Federation, log-based CDC (Debezium), and modern Flink use cases.
Staff Engineer (Data Platform) · Grupo SBF
Led the evolution of the data platform at Grupo SBF (Nike Brasil, Centauro). Designed and built a streaming data platform for low-latency real-time analytics on Apache Beam, Kafka, Schema Registry, and BigQuery. The platform enabled real-time pricing, reducing inventory overhead and corporate debt. Cut Apache Beam / Dataflow cloud cost by 45%. Architected a Kubernetes sidecar that reduced new-producer onboarding from ~80 hours to under 4.
Platform Software Engineer · Zwift
Server platform team at the e-sports unicorn. Zwift's backend is an event-driven system with sub-second client interactions and hundreds of thousands of concurrent users. Designed and developed the algorithms behind the real-time leaderboards — eliminated the platform's performance bottleneck and unlocked 300K concurrent users. Extracted services from the monolith and built platform libraries.
Co-founder & CTO · Symprise
Technical and product leadership across 20+ enterprise software projects for Petrobras (largest Brazilian company), Petroflex, SEARA, General Electric and Saint-Gobain (Brazilian operations), and others. Led Petrobras' first 3-tier transactional e-business: self-service refinery-delivery scheduling. Led development of GlobalTrader® WebWorkPlace® — the first Brazilian web platform for international-trade business processes, with full domain coverage and inter-organizational collaboration. Architect and built declarative frameworks for fast development of Business Systems. Pioneered (Brazil) direct SAP R/3 integration via SAP Java Connector and BAPIs (2002).
Researcher · NCE / UFRJ
Computer Engineering R&D at the Federal University of Rio de Janeiro. Team member on Plurix — a Unix-like multi-threaded operating system for shared-memory multiprocessors (1984–1987, POSIX-style multi-threading in 1986). Built Plurix Shell, Make, etc, and the Plurix Mac+ macro-processor. Wrote the Z80 assembly firmware for NCE's graphical color terminal, presented at the Brazilian Society of Computing Congress.
Recent Open-source contributions
awslabs/fullstack-solution-template-for-agentcore— AWS reference Agentcore agentic systemaws-samples/sample-strands-agent-with-agentcore— AWS reference Strands/Bedrock/AgentCore agentic systemspring-ai-community/spring-ai-agentcoreSYSTRAN/faster-whisper - Speech to Text.
sivaprasadreddy/sivalabs-agent-skills— Spring Boot AI Skillwundergraph/cosmo— performant, cost-effective GraphQL Federation implementation
Certifications (selected)
- AWS Generative AI and AI Agents with Amazon Bedrock (Coursera/AWS, Jan 2026)
- Building Production-Ready AI Agents with Amazon Bedrock AgentCore (AWS Training, Apr 2026)
- Architecting AI Solutions — Scalable GenAI Systems (Coursera, Oct 2025)
- Claude Code: Software Engineering with Generative AI Agents (Vanderbilt University, Sep 2025)
- Model Context Protocol — Introduction and Advanced Topics (Anthropic, Sep 2025)
Education
- PhD Coursework, Business Management — COPPEAD UFRJ
- M.Sc., Computer & Systems Engineering — COPPE Sistemas, UFRJ
- B.Sc., Computer Science — UFRJ
Selected publications
- Distributed Dynamic Scheduling for High-Performance Multicomputers — UFRJ COPPE Sistemas, 1992
- A Communication System for Large Scale Multicomputers — SBAC-PAD V, 1993
- MULPLIX: A Unix-like Multi-Threaded Operating System for MULTIPLUS — SBAC-PAD III, 1990
- Primitives for Parallel Programming on MULTIPLUS — SBAC-PAD V, 1993