Partner at Lux Capital
Check size: No personal check-size range is published. Lux's USD 7B under management and disclosed company rounds are not Grace's personal range; she is explicitly described as a first-check and early investor.
Grace backs computational-science companies where software or AI transforms difficult physical, institutional, or technical systems. She looks for first-check opportunities in infrastructure and applications, including model platforms, developer tools, generative media, robotics, and AI-native enterprise workflows. Her product lens starts with a painful real-world experience and asks whether new fundamental technology enables an excellent end-to-end product, rather than an AI feature bolted onto an old workflow.
Begin with the specific experience that 'sucks,' quantify its cost, and show how the technology makes a redesigned end-to-end product possible. Demonstrate production reliability, workflow integration, proprietary data or technical advantage, model and compute economics, user outcomes, and why this team can become an early category leader. For infrastructure, show developer love and application pull; for agents, show successful task completion, not just conversational quality.
Exceptional technical teams, a large frustrating workflow, a fundamental technology shift, first-check opportunities, resourceful and thoughtful founders, products that work reliably in production, model-agnostic or infrastructure leverage, global ambition, and companies that bridge software with the physical world.
Her writing suggests weaker fit for thin chatbot wrappers, AI added without redesigning the workflow, demos that cannot operate reliably in enterprise production, products without measurable user value, consensus opportunities lacking a differentiated technical insight, and infrastructure without a clear developer or application pull. These are thesis-derived signals, not a formal pass list.
Start with a painful, repeated real-world experience; identify the new technical primitive; then redesign the full workflow around a materially better outcome instead of appending an AI feature.
Measure autonomous task completion, resolution quality, latency, cost, human escalation, integration depth, security, compliance, multilingual performance, and reliability across real enterprise edge cases.
Validate that developers adopt, retain, and build valuable applications on the platform and that abstraction expands access without imposing lock-in or a painful graduation path.
For AI touching biology, robotics, defense, energy, or other physical systems, connect model performance to sensing, actuation, safety, deployment constraints, and measurable real-world outcomes.
New York's concentrated talent and cross-industry density make it a durable center for AI applications, infrastructure, and scientific computing companies.
Rapidly improving and diversifying models create opportunities for AI agents and product experiences that move beyond consensus chatbot patterns.
AI should redesign an entire frustrating workflow and deliver reliable, integrated outcomes in production rather than merely add a chatbot to the existing process.
Developer infrastructure can expand a market by removing the need to master fragmented stacks while avoiding the limitations and graduation risks of low-code tools.
+ 38 more investments. View fund →