Maylla Souza

Built WhatsApp conversational AI agents for SMBs, turning people-dependent operational flows (FAQs, scheduling, lead intake, follow-ups, urgent requests) into modular multi-agent architectures with clear scope and handoff rules. Prototyped conversational flows directly with LLMs on low-code platforms for fast validation, then implemented safety layers that route sensitive/urgent cases to humans. Created structured test scripts and worked with model-agnostic, cost-aware integrations for tools without strong APIs. Self-hosted the stack using Docker and cloud infrastructure, monitored deployments, and packaged the solution as a product (briefing, proposal, contract, pitch deck, project board).

Project-based product design for B2B and fintech teams, including CRM/ERP redesign, design systems, UX audits, QA, and clickable prototypes. Delivered a multi-track engagement for a multi-franchise Field CRM that included live ERP CRM redesign/QA alongside a greenfield platform proposal. Tested multiple functional journeys in staging, versioned a backlog, and built a shared Material-based design system across products. Pivoted deliverables based on stakeholder feedback and produced a board-ready platform thesis, plus clickable prototypes and QA journey outputs.

Design and QA across a prediction-market platform with white-label player app, admin/back-office suite, and partner embed widgets. Converted PRDs into player/admin/back-office flows, covering the full event and market lifecycle (create, approve, suspend, close, cancel, dispute and resolve) and multi-tenant operations. Owned the company design system end to end (architecture, tokens, i18n, versioned releases) and led zero-breakage migrations for 280+ component instances; drove adoption above 90%. Led QA using Gherkin test execution, exploratory testing, browser automation, and bug triage (including release-blocking issues with strong evidence). Wrote test plans for new payment rails and established release rituals and governance.