About
The forge and the smith
“A thousand days of training to develop, ten thousand days of training to polish.”
Miyamoto Musashi
I'm Kayo Souza (Kinho), a senior software engineer, tech lead and AI engineer with 8+ years in production, 4 of them with US clients. From Vitória, Brazil, I've worked for Nike (the Asset Library behind 16M+ assets, built as sole engineer, and AutoCrop, a GenAI service that cut designers' prep time 44%), QuintoAndar, Loft and Yooga, where I led the 5-engineer payments team behind US$ 70M+ processed in 6M+ transactions.
Today I'm Tech Lead of the nearshore team on Care Agent, the LLM patient agent of Experity, which holds half of the US telemedicine market: booking runs in code instead of the model, PHI stays out of every telemetry event, and the agent runtimes run under a 99.99% availability SLO, all under HIPAA. I co-founded TOPO Contábil, where I built the architecture, engineering and security; it reached five-figure ARR in its first year. I'm also a software engineer on a research grant, across all tracks, in the modernization project of the Espírito Santo Court of Justice (TJ-ES). Top 0.01% Claude Code user globally.
Kaizen Forge is the software house I founded to bring that standard to other companies: custom software, AI consulting and dedicated developers, built by a team that uses AI as a tool and by a tech lead who is accountable for every delivery. Kaizen (改善) is continuous improvement; a forge is where iron becomes blade under pressure and heat. That's how we treat software and AI: no hype, made to measure, measured.
I write about engineering and AI at kinho.dev.
Where the standard was forged
- Nike
- QuintoAndar
- Loft
- Yooga
- Experity
Principles
Honesty over polish
A number that wasn't measured doesn't go on the site. A scope that doesn't close doesn't get promised.
Production or nothing
Demos don't pay bills. Every project ends with a system operating and a team trained.
Kaizen
Metrics-driven continuous improvement. Every system I ship gets better, and cheaper, over time.
The manifesto
Kaizen is not a slogan
Every AI company promises overnight transformation. We don't. Kaizen (改善) is continuous improvement: small, measured, compounding. A system a little better every week beats a revolution that never leaves the slide deck.
We forge it in this order, every time: diagnose with honesty, cut what doesn't pay for the project, ship to production with metrics from day one, and keep polishing. It's not a magic formula. It's discipline, applied again and again.
If you want a nice case study for a pitch deck, we're not for you. If you want software that runs, built by a team that proves it with numbers, that's exactly what Kaizen Forge exists for.
Cut the hype. Measure what matters. Repeat.
The squad
The team behind the code
A lean team from Vitória, Brazil, with AI at the center of the process: the people who sell, build, and respond are the same team, with no layers between you and delivery.
Gabriel Wagnitz
Head of Security
Computer Science (UFES), then most of a decade in Canadian cybersecurity: built and ran SOC ingest processing 3+ TB of security telemetry per day for 80+ organizations, with 10+ incident-response engagements. Also the co-builder of an offline-first civic platform he still maintains, seven years on. One engineer who builds the feature and secures the infrastructure under it.
SecuritySIEMReact
João Holz
Solutions Architect · ML & Data
A statistician turned architect: time-series forecasting, production ML systems and data platforms. Architected Oráculo end to end, delivered a public-sector data platform serving 300+ municipalities leading a 6-12 engineer team, and shipped the MLOps architecture for a payment gateway. Research exchange in computer vision at Brno University of Technology.
ArchitectureGoPython
Lucas Fabris
Lead AI Engineer
Built an AI lead-scoring product from an empty repo to production as sole engineer. Developed a native Android POS to stable release in two months, and is a key player in the payment integrations (Tuna, Apple Pay, Google Pay) behind US$ 70M+ processed. Owner attitude: takes things end to end.
AI EngineeringNestJSAndroid
Ana Có
Designer · Lead Front-end
A designer by degree who engineers: platform team at QuintoAndar, now owner of the design system behind a merchant app serving 4,393 active stores: 841 commits of components, prototypes and product screens. The reason our work looks like product, not project.
Design SystemsTypeScript · ReactAngular
Eden Kiefer Junior
Full-stack Engineer
Backend-focused full-stack engineer, twice dropped into stalled projects and twice took them to production: a cross-border e-commerce live in one month, a municipal accounting system in three, as tech lead. Also owns the loyalty vertical on an engine with 1.1M participating consumers. Speaks little, ships a lot.
Full-stackJavaIntegrations
Willian Kuhn Littig
Automation & Infrastructure Engineer
Builds the operational backbone: marketplace integrations processing 1,000+ invoices a month, WhatsApp automations where AI answers questions on its own, and Oráculo's persistence backend: 10,000+ real sessions processed in production.
AutomationInfraNode
Gustavo Stein dos Santos
DevOps & Backend Engineer
Backend and infrastructure: built the browser SDK and Go ingestion API feeding Oráculo's models (10,000+ production sessions), stood up a BI department's infrastructure from zero and led its migration to AWS: 300+ dashboards serving millions of rows with no downtime, on a full Grafana observability stack.
DevOpsGoData
Ana Carollyne
Developer · Web & AI
Builds client sites and AI integrations into the workflow, from briefing to launch.
TypeScriptReactIA
Matteo Pozzatti Ciardo
Sales & Relationship
Handles prospecting, support, and client follow-up, from first contact to after-sales.
VendasRelacionamentoClientes
Ricardo Secchin
Lead Mobile
Android developer with apps in production, who has worked on teams in Brazil and Portugal. Leads the team's mobile work.
AndroidMobile
Ready to cut the hype?
A free 30-minute assessment. Tell us what you need built: you'll leave knowing where software generates return in your company and what it takes to deliver it, deciding on evidence, not opinion.