Participated · Seed · Aug 2026

Lemma builds production monitoring infrastructure for autonomous AI agents by capturing structured traces of agent executions, including LLM calls, tool usage, inputs and outputs, retrieval steps, timing data, and errors. Its platform analyzes those traces to identify recurring semantic failures, prioritize issues, and surface alerts. The company aims to detect silent failures where agents appear to succeed technically but produce incorrect outcomes. Lemma was founded by Jerry Zhang and Cole Gawin and participated in Y Combinator’s Fall 2025 cohort. It recently raised $2.3 million in a pre-seed round to accelerate development of its monitoring and failure-detection tools. Initially, Lemma plans to focus on supporting startups that are already running AI agents in production.

Participated · Series A · Jul 2026

Weave offers an engineering intelligence platform that analyzes pull requests, code reviews, deployments, and AI-generated contributions to produce a unified measurement of engineering output. The company uses machine learning models to connect coding activity with business progress and to estimate the AI spending required to produce an hour of completed engineering work. Features include recommendations for lower-cost AI models based on production data and metrics that surface AI adoption, engineering quality, and cost of AI-generated work. Weave reports having analyzed more than 2 million contributions across over 20,000 engineers at more than 500 organizations, and counts Robinhood, Reducto, and PostHog among its customers. Based in San Francisco, the startup aims to curb what it calls “tokenmaxxing” by shifting focus from volume-based metrics to outcome-based measurements. The company will use the Series A proceeds to accelerate product development and expand its platform.

Participated · Seed · May 2026

Modiqo builds an infrastructure layer for reliable AI agent workflows and has launched a product called Rote that captures successful agent executions and turns them into deterministic, repeatable processes. Rote is intended to make AI work consistently across changing models and APIs, reduce token usage by reusing proven workflows, integrate with existing tools without heavy engineering, and provide visibility into executions and costs. The company positions its offering as an execution layer to help teams move AI projects from experiments to production-grade systems. Modiqo states its target users are organizations deploying AI across real-world operations that need reproducibility, reliability, and cost control. The company announced a $3 million pre-seed raise to support Rote's launch. The press release was issued from San Francisco on May 28, 2026.

Participated · Seed · Oct 2025

Popsink has built a Change Data Capture (CDC) solution that continuously streams changes from legacy sources—such as ERP, CRM, mainframes, and Oracle/IBM databases—and replicates them in real time to modern environments like Snowflake, Databricks, and Iceberg. By removing the need for heavy data migrations, the platform offers enterprises a secure, low-risk path to feed analytics, business intelligence, and generative AI projects with live operational data. Co-founded by Benjamin Djidi, a former Amazon and Sixt engineering leader, and CTO Alexandre Colella, Popsink positions itself as the critical ‘invisible bridge’ between operational cores and intelligence layers. The technology targets companies struggling to modernize data stacks amid growing pressure for real-time AI readiness and data sovereignty. Following strong demand for real-time infrastructure, the startup plans to double its workforce, accelerate product development, and extend its market presence across Europe and North America. No revenue or user figures were disclosed in the article.

Participated · Series A · Sep 2025

Founded in 2020, Paris-based Qovery offers a SaaS platform that automates the entire lifecycle of application deployment and infrastructure management across public clouds and on-prem Kubernetes clusters. The product installs in about five minutes and is positioned as delivering the output of four to five DevOps engineers, helping customers cut setup times from weeks or months to roughly a day. Users report roughly 3× gains in DevOps efficiency and can achieve hundreds to thousands of deployments per day while maintaining compliance and security. By abstracting away cloud complexity and avoiding vendor lock-in, Qovery allows developers to focus on coding rather than infrastructure management. The company monetises its platform as a subscription service and has attracted customers seeking faster, lower-cost software delivery. With fresh funding, Qovery plans to expand geographically, grow its team, and invest further in AI-driven product capabilities. To date, the company has raised a $13 million Series A round in addition to prior seed capital.