Participated · Seed · Jul 2026

Walden Robotics develops and deploys general-purpose robots powered by Large Behavior Models and other Physical AI advances to perform useful work in manufacturing, logistics, and other industries. Founded in 2026 and launched out of Toyota Research Institute, the company combines foundational robotics research with in-house operations teams to bring robots into production environments. Walden reports that its robots have been operating in production at a Toyota plant in North America since February, progressing from pilot to real work in under two months. The company has strategic partnerships across automotive, aerospace, semiconductors, electronics, logistics, and life sciences and emphasizes human-centered deployment where robots work side-by-side with people. Walden’s stated plans center on scaling commercial deployments and continuous on-the-job learning to improve capabilities over time. Financially, the company announced a $300 million seed round at a $1.1 billion valuation led by Toyota and Deviation Capital, providing capital to support its launch and customer expansion.

Participated · Seed · May 2026

Tensormesh builds and sells Tensormesh Inference, a SaaS inference platform that uses KV caching to store and reuse computed model state to avoid redundant GPU computation. The platform delivers up to 10x reductions in latency and GPU spend by serving repeated context from cache and reports cache hit rates, token-level cost breakdowns, and other performance metrics in real time. Tensormesh offers serverless, OpenAI-compatible API access for immediate use and reserved deployments for enterprises needing dedicated capacity and SLAs. The company also maintains and contributes to the open-source LMCache project (over 8,000 GitHub stars) and highlights integrations across vLLM, SGLang, TensorRT, and multiple cloud vendor offerings. Tensormesh is led by CEO Junchen Jiang and was founded by faculty, PhD researchers, and alumni from the University of Chicago, UC Berkeley, and Carnegie Mellon. The company has raised $24.5 million in total funding and markets a pricing model that bills cached input tokens at $0 for serverless deployments, with well-optimized customers regularly achieving cache hit rates above 70%.

Participated · Seed · Mar 2026

Standard Kernel builds an AI-driven platform that automatically produces instruction-level, hardware-specialized GPU kernels, replacing static libraries with code precisely tuned to each workload and accelerator. In partner tests on NVIDIA H100 GPUs, the software has delivered 80 % to 4× end-to-end performance gains, occasionally surpassing NVIDIA’s own cuDNN. By operating deep in the systems stack, the company aims to provide day-one peak efficiency for every new chip without months of manual tuning. The Palo Alto–based team includes alumni from MIT, Stanford, UIUC, and SJTU, and has created open-source projects such as KernelBench and Kernel Tree Search. Newly raised capital will fund continued R&D, expansion of deployments with AI-native and enterprise customers, and progress toward adaptive systems software that evolves alongside emerging models and hardware. The company is actively hiring engineers and researchers to advance its kernel generation technology.

Participated · Equity · Sep 2025

Salt AI offers an adaptive, SaaS-based collaborative operating system built for life-science and health-tech enterprises. The platform allows research and clinical teams to rapidly integrate, visualize, and operationalize AI models for tasks such as protein generation, biomarker discovery, clinical evidence synthesis, and diagnostics. With a visual-first interface and enterprise-grade scalability, the product is positioned to shorten R&D cycles and improve cross-functional collaboration in biotech and pharma. The company recently bolstered its leadership with veterans from Boston Consulting Group, Doctor Evidence, Concert AI, and Thermo Fisher Scientific, underscoring its commitment to pair scientific expertise with advanced machine learning. Following a $10 million funding round, Salt AI is scaling its adaptive pipeline across the life-science value chain and preparing demonstrations—such as oncology-focused protein design—at industry events like the 2025 AI Drug Discovery & Development Summit. CEO and co-founder Aber Whitcomb leads the firm’s push to deliver practical AI impact for translational scientists and drug developers. Financial details beyond the latest $10 million infusion, including revenue or user metrics, have not been disclosed.

Participated · Equity · Jul 2025

Moonvalley builds foundational AI video models and a production‑grade generative videography platform, Marey, aimed at professional filmmakers and enterprise brands. Marey was recently released to the public via Moonvalley.com and the company works with its filmmaking arm, Asteria, to deliver cinematic, commercial‑ready outputs. Moonvalley emphasizes models trained only on licensed content to address IP and studio concerns. The company raised an additional $84 million to scale operations and meet enterprise demand, bringing total funding to $154 million. Planned uses for the new capital include expanding its licensed content library, developing API access for developers, building studio‑requested features, and growing engineering and support teams for enterprise deployments. Moonvalley develops AI video-generation tools and models (notably its Marey model) that produce up to 30-second HD clips with fine-grained camera and motion controls. The company emphasizes data provenance and legal risk mitigation, packaging and purchasing licensed videos for training rather than relying solely on public data. Moonvalley is building a user-facing interface with storyboarding and granular clip-adjustment tools that it has not yet previewed publicly. Marey can generate video from text prompts, sketches, photos, and other clips, and the company is collaborating with an AI animation studio called Asteria. Moonvalley says it will allow creators to request content removal, let customers delete their data, offer an indemnity policy, and block disallowed content and likeness-based prompts. The startup is positioning these safeguards and curated training data as its primary differentiation in an increasingly crowded generative-video market. Moonvalley is a Los Angeles–based deep learning research company focused on building next-generation generative AI creative models and an artist-first creative suite for the film industry. The company develops creative tools and deep learning models aimed at enhancing trust, transparency, and efficiency in the media industry. Moonvalley says it works with researchers, engineers and artists from DeepMind, Google, Meta, TikTok, Microsoft and more to design video and creative AI solutions. The team is led by co-founder and CEO Naeem Talukdar alongside John Thomas, Mateusz Malinowski, Bryn Mooser and Mikolaj Binkowski. Moonvalley plans to use new capital to continue advancing AI research, expand R&D, hire top talent, and accelerate deployment of its solutions. The company positions itself as a trusted partner for media organizations while pursuing frontier foundational video models.