Marcos Kuchak Filho

• Architected and built a framework-agnostic AI agent platform from scratch, consuming five SQS event queues and running 13+ production reactive and proactive workflows automating mortgage-loan processing (daily risk briefings, borrower and lender communication, document recovery), with confidence gates executing high-confidence actions autonomously and routing lower-confidence work to processors for approval, under full audit logging.
• Designed and shipped an OAuth 2.1-protected MCP gateway exposing ~90 loan-data tools to AI agents (Claude, GPT, Cursor), standardizing tool schemas to cut the agent tool-call failure rate from 44% to under 10%.
• Engineered layered LLM cost controls (deterministic pre-filters, risk fingerprinting, and per-workflow multi-provider model selection), cutting the highest-spend workflow's LLM cost by 80-85%.
• Built the platform's observability and quality layer: per-workflow tracing with full cost breakdown, LLM eval suites (deterministic and LLM-as-judge), a React analytics dashboard, Slack failure alerting, and a scheduled worker detecting silently-missed runs.
• Designed, built, and launched an event-driven audit-logging microservice in one week (~19.5K LOC) with field-level AES-256-GCM PII encryption, then integrated it across five services and the web app.
• Built instant search end-to-end with Algolia: secured per-tenant API keys, 18 virtual replicas for sorted queries, idempotent bulk and real-time message-bus sync, and automatic MongoDB fallback, cutting search time on any loan field from 3-10s to 50-100ms.

Tech stack: TypeScript, Node.js, Remix/React Router, Express, Zod, MongoDB (CSFLE), Redis, AWS SQS, Model Context Protocol, Anthropic/OpenAI/Gemini APIs, Mastra, OpenAI Embeddings & Vector Search, Algolia, Terraform, AWS ECS Fargate, Kubernetes, Docker, CircleCI, OAuth 2.1 (Clerk)

• Led development of a scalable full-stack system in a TypeScript monorepo, leveraging React and Next.js for a dynamic front-end alongside Node.js, Fastify, and tRPC with hexagonal architecture for robust APIs, which reduced code duplication by 30% and accelerated feature delivery by 40%.
• Implemented a secure, passwordless authentication system with OTP, reducing login friction and boosting user retention by 25%.
• Designed and implemented a generative AI-powered conversational BI tool that renders live data visualizations directly inside the chat interface, letting farmers query field operations in natural language.
• Developed a provider-agnostic AI agent framework with pluggable model adapters (OpenAI, Anthropic, Gemini), unifying horizontal business domains with specialized vertical agents and enabling seamless agent handoffs with tool use in the background.

Tech stack: TypeScript, React, Next.js, Remix, Node.js, Fastify, tRPC, Prisma, PostgreSQL, Redis, Turborepo, Tailwind CSS, LangChain, Python, Docker, GitHub Actions, AWS, Cloudflare, DigitalOcean

• Enhanced a law firm's prospect capture page, integrating a responsive Vue.js front-end with Node.js back-end endpoints for data validation, increasing form fill rates by 40-50% and driving lead generation.
• Developed a web-based Business Intelligence app with a React front-end and a NestJS back-end, turning raw client data into live dashboards and reports.
• Refactored a health plans portal, implementing a Vue.js front-end for improved user experience and integrating a Node.js back-end for chatbot logic, which enhanced sales team effectiveness and interview accuracy.

Tech stack: TypeScript, React, Vue.js, React Native, Expo, NestJS, Express, Node.js, GraphQL, PostgreSQL, MySQL, MongoDB, Prisma, Redis, RabbitMQ, Jest, Cypress, Docker, GitHub Actions, AWS

• Designed and built a back-office platform, leveraging React for the user interface and Node.js for the back-end, significantly increasing content generation capabilities and improving workflow efficiency by 60%.
• Optimized SEO for content delivered through React applications, leading to over 1,000,000 users in a few weeks through improved search rankings and discoverability.
• Implemented a notification system, increasing organic traffic by 30-35% during off-peak hours.
• Developed a marketing campaign management tool leveraging Node.js for automated ad control (Facebook/Google Ads APIs), adjusting ad spend in real time based on ROI.

Tech stack: React, Vue.js, Node.js, GraphQL, MySQL, Python, Bash, OneSignal, Tailwind CSS, Jest, AWS

• Managed end-to-end projects from ideation to execution, consistently delivering 100% on time.
• Developed custom websites and e-commerce platforms, leveraging PHP and Laravel for scalable back-ends, and specialized platforms like WordPress and Magento.
• Engineered bespoke web applications using Node.js for robust back-end development, ensuring high performance and scalability for client-specific needs.

Tech stack: Node.js, PHP, Laravel, WordPress, Magento, MySQL, jQuery, Cloudflare