Case study · Conversational AI · B2B SaaS
Commercial WhatsApp operations SaaS (conversatio.app) featuring multi-agent inbox and AI triage. Technical leadership focused on quality engineering and automated AI testing: synthetic conversation eval pipelines, prompt regression testing, end-to-end chat flow validation, and webhook concurrency simulation.
Build a platform fully integrated with WhatsApp Business to send messages to customers and to manage and share conversations across sales and support teams, with multiple agents.
Built an automated evaluation pipeline (AI Evals) using synthetic conversation datasets to systematically measure intent classification accuracy, urgency detection, and entity extraction precision across model prompt updates prior to production deploys.
Engineered automated regression test suites to validate generative response consistency and strict JSON Schema outputs, protecting the platform against contract breaks, prompt drift, and hallucinated handoffs in edge-case buyer dialogues.
Implemented E2E test suites emulating concurrent operator sessions and inbound WhatsApp Business API webhook spikes, verifying message idempotency, queue dispatch integrity, and zero duplication under high concurrency.
Led engineering best practices across the development team, establishing mandatory test coverage on PRs, automated quality reports, and CI/CD deployment gates to ensure stable, release-ready software with every sprint.
Measured on September 23, 2026 at conversatio.app, worst of two runs, with a phone profile: 150 ms latency, 1.6 Mbps and a 4× slower CPU. The platform itself is still in development.