AI Product Engineer at Elephant (January 2026 - ongoing)
Built an automatic bug-fixing AI pipeline that triggers from Intercom, collects bug context from Intercom and internal logs, and creates a pull request with an implemented fix. Reworked an AI course generation workflow from low-quality LLM behavior into a higher-quality AI agent. The work focused on improving model behavior through structured agent workflows rather than manual data annotation. • Triggered pipeline via Intercom events and enriched prompts with operational context • Generated code changes automatically by turning bug context into actionable fixes • Refactored LLM workflow into an agentic system for better course generation outputs • Implemented evaluation/quality improvements through iterative agent behavior testing