Stop Burning Tokens: Why self-improvement needs domain expertise first - Annabell Schäfer, Langfuse
AI Engineer World's Fair 2026 · 17:39
AI engineering, LLM observability and evaluation
Langfuse builds an open-source AI engineering platform for teams developing LLM applications and agents. It combines tracing, prompt management and evaluation so developers can inspect individual model calls and tool executions, test changes through experiments, and monitor production cost, latency and quality. Teams can update prompts without redeploying their applications, run Python or TypeScript checks, and use LLM-as-a-judge evaluations. The platform is available through Langfuse Cloud or self-hosting.
Co-founders Marc Klingen, CEO, Max Deichmann, CTO, and Clemens Rawert, COO, began working together in 2022 and publicly launched Langfuse in 2023 after developing AI applications during Y Combinator. Its engineering has focused on making production traces practical to analyze at scale. A move from Postgres to ClickHouse supported analytical workloads; the subsequent observations model stores agent steps in an immutable, denormalized table, removing query-time joins and deduplication.
ClickHouse completed its acquisition of Langfuse in January 2026, with the whole team joining and cloud and self-hosted offerings continuing. Langfuse had raised a $4 million seed round in 2023.
AI Engineer World's Fair 2026 · 17:39
AI Engineer Europe 2026 · 24:09
Affiliations reflect their AIE appearances, not necessarily current employment.