led the frontend, backend architecture and AI development of a learning platform serving 50,000+ learners across South America, Asia, Israel and England, taking features from spec to production alongside product, design and business stakeholders.

• built and owned the complete learning client using React and TypeScript, including adaptive assessments, reading and comprehension flows and media-rich lesson viewers
• led personalisation work that lifted engagement by 35% and retention by 25%
• built agentic flows and automations that help children practise reading, writing and comprehension
• designed RAG pipelines with hybrid retrieval, re-ranking, cited answers and knowledge graphs, tuned against evals for accuracy, latency and cost
• shipped multi-agent systems to production using tool calling and MCP, with schema validation, prompt-injection filtering, human approval gates and per-run tracing
• streamed LLM answers over SSE from Bedrock, Anthropic and OpenAI, with structured outputs and cross-provider fallback
• built a Manifest V3 Chrome extension that brings the AI assistant into the page the user already has open
• built backend services using Node.js (Express, NestJS) and Python (FastAPI) on PostgreSQL, MongoDB and Redis, with versioned REST APIs and Pytest coverage
• integrated Stripe billing, Google Workspace and Monday.com CRM behind OAuth and SSO, cutting operational overhead by 30% and billing support tickets by more than half
• migrated deploys to AWS (ECS, Fargate, Lambda, S3, SQS) using Terraform and GitHub Actions, cutting deploy time by 40%