Participated · Seed · Jul 2026

Runta develops an execution layer for AI agents that puts hard boundaries around what an agent can reach and spend while keeping a replayable record of what it executed. The company positions agents as a new kind of computer user and emphasizes combining permissions, identity, spend limits and auditability into a single infrastructure layer. Founder Guanlan Dai announced Runta's public launch alongside a $20M seed raise and invited teams to apply for early access. Runta's public messaging highlights its goal to enable production use of agents by providing control and accountability that traditional software infrastructure lacks. The company did not disclose revenue, user metrics, founding year, or location in the announcement.

Participated · Series A · Jul 2026

Applied Computing’s core product, Orbital, combines time-series modeling, physics- and chemistry-aware simulation, and language modeling to analyze sensor readings and predict the state of industrial facilities. The platform is designed to flag anomalies, investigate root causes, and simulate how changes in one part of a facility could affect other operations, with the company claiming it can compress multiday investigations into seconds. Founded in 2023, the company has grown quickly and reports double-digit millions in annual recurring revenue within 18 months of emerging from stealth. Orbital is deployed at several large, publicly listed oil, gas, refining and petrochemical customers and is integrated into partners’ platforms such as KBR’s INSITE 3.0; Applied Computing also partners with Wipro and is working with a major U.S. upstream operator. The company has headquarters in London, an operational hub in Bengaluru, and recently opened an office in Houston as it pursues international expansion and additional deployments.

Participated · Series A · Jun 2026

Tsuga builds an observability platform that runs directly in customers' cloud accounts (AWS, Azure, Google Cloud and regional sovereign clouds) so telemetry data stays inside the customer's environment rather than being routed through a vendor. The company offers a single billing rate per gigabyte, claims to provide unsampled, end-to-end traces (including AI agent traces, prompt and token details, confidence scores, and agent call graphs), and assists customers in optimizing setups to reduce data usage. Founded in 2024 by Gabriel-James Safar and Sébastien Deprez (who previously sold Madumbo to Datadog), Tsuga exited stealth and within six months reported several million dollars in revenue and secured enterprise customers such as Le Monde, Camunda, Buk, and Black Forest Labs, with average contract values in the six figures. Tsuga's product targets cost and data-sovereignty challenges in modern AI-heavy workloads by keeping data inside clients' security perimeters. The company plans to use recent funding to expand its business and speed development of its AI agent observability capabilities.

Participated · Series B · May 2026

OpenRouter operates as middleware between developers and more than 300 open-source and proprietary large language models, offering a unified API that lets developers compare quality, price, and latency and dynamically route tasks to different models. The platform charges a percentage service fee when developers purchase credits to run models, generating revenue that the article reports has grown from about $10 million annualized in October to over $50 million annualized currently. Founded in 2023, the company enables multi-model architectures by allowing applications to call different models for different task types and supports developers in optimizing cost and performance. The article highlights OpenRouter's neutral positioning and broad model coverage as competitive advantages versus cloud providers and startups that may favor their own models. It faces competition from cloud giants offering multi-model access and startups like PortKey, and must manage neutrality and supplier relationships as it scales. The company is positioned as addressing developer demand for simpler multi-model orchestration amid growing AI application complexity.

Participated · Series E · May 2026

Sigma is built on the cloud data warehouse and provides a governed workspace that supports a spreadsheet interface, SQL, Python, and native AI to let teams explore live data, build applications, and automate workflows without moving data. The company emphasizes secure, warehouse-native architecture and has launched products including Sigma Agents (no-code agents), Sigma Assistant (an AI copilot), Sigma Data Modeling Skills for AI Agents, and the Sigma MCP Server for governed answers into AI chat assistants. Sigma reported strong financial momentum, surpassing $200 million in ARR in April 2026, achieving 100%+ year-over-year growth in the latest fiscal year, and adding over 1.1 million new active users in that period. It serves more than 2,000 customers worldwide, including large enterprises and AI innovators. The announcement originated from Sigma's San Francisco office and framed the company as driving adoption of agentic analytics and AI workflows while emphasizing governance, reliability, and security.