Aditya Kulkarni

About

LLM-focused engineer with end-to-end experience shipping an OpenAI-powered edtech teacher assistant into production, using Humanloop-driven prompt iteration, rigorous observability (Datadog), and A/B testing tied to real learning metrics (25% comprehension lift). Also led adoption-driving technical demos at SiriusXM (event-driven AWS Lambda/Kotlin/CDK pipeline cutting processing from 24 hours to seconds) and partnered with sales at Spresso.ai to close eCommerce SDK deals and boost activation from 40% to 85%.

Experience

Education

Key Strengths

• Took an OpenAI-powered edtech assistant from prototype to production with human-in-the-loop safety
• Data-driven iteration using weekly user feedback loops + A/B tests tied to acceptance and learning outcomes
• Improved student comprehension by 25% through streaming + interactive charts; maintained production-grade reliability
• Rapid incident diagnosis and remediation in production (35% rejection to 92% acceptance in 48 hours via context-truncation fix)
• Strong observability-first approach (prompt/I-O logging, latency/token/error metrics in Datadog) with canarying and model/cached fallbacks
• Led highly actionable technical demos that drove org-wide adoption (pipeline processing reduced from 24 hours to seconds; eliminated 5 hours/week manual work)
• Effective cross-functional partnership with sales/client success to close deals and drive adoption (SDK activation 40% to 85%; addressed Shopify integration concerns)
• Customer-driven feature prioritization using structured impact scoring and usage metrics (latency cut 60% to retain at-risk customer)