Portfolio

Infrastructure for deploying and serving AI models in production.

Thesis connection

Model capability creates value only when it can be delivered reliably. Inference infrastructure connects model choice, serving economics, and the demands of real applications.

Thematic fit, not the original investment rationale.

What matters in this area

Serving performance, operational reliability, and cost at the application’s actual workload.

Multimodal data infrastructure for storing, searching, and working with AI data.

Thesis connection

AI applications need more than a place to store vectors. A shared foundation for multimodal data can make retrieval and data-intensive AI workflows easier to build and operate.

Thematic fit, not the original investment rationale.

What matters in this area

Data access patterns, retrieval quality, and the path from experimentation to production.

Kafka-compatible streaming built around a diskless architecture.

Thesis connection

Fresh information depends on reliable data movement. Simpler streaming infrastructure can reduce the operational burden beneath analytics and intelligent applications.

Thematic fit, not the original investment rationale.

What matters in this area

Compatibility, system behavior under load, and infrastructure economics.

A database bringing transactional and analytical workloads into one system.

Thesis connection

Separate operational and analytical systems introduce data movement and coordination costs. Database architectures that narrow that gap can support faster, more contextual applications.

Thematic fit, not the original investment rationale.

What matters in this area

Mixed-workload performance, consistency, and the practical cost of adoption.

Real-time feature computation and data infrastructure for machine learning.

Thesis connection

Model quality depends on the data available at decision time. Fresh, consistent features connect training assumptions with the reality of production inference.

Thematic fit, not the original investment rationale.

What matters in this area

Feature freshness, consistency across workflows, and developer control.

Web search and data access for AI agents.

Thesis connection

Useful agents need current external information. Reliable access to web data makes freshness and relevance part of the infrastructure rather than a manual collection task.

Thematic fit, not the original investment rationale.

What matters in this area

Coverage, data quality, and suitability for the agent’s task.

An extensible MySQL distribution for the agentic AI era.

Thesis connection

Existing databases remain central to enterprise software. Extensibility offers a path to new capabilities without requiring every organization to replace its data foundation.

Thematic fit, not the original investment rationale.

What matters in this area

Compatibility, extension boundaries, and upgrade safety.

A semantic layer that gives AI the business context behind enterprise data.

Thesis connection

Access to data is not the same as understanding it. Shared definitions and relationships can make business context a reusable foundation for AI-driven analysis.

Thematic fit, not the original investment rationale.

What matters in this area

Definition quality, governance, and how business meaning stays current.

Knowledge infrastructure that captures and maintains context for software teams and agents.

Thesis connection

Software knowledge is distributed across code and conversations. Reusable, maintained context can reduce the cost of repeatedly reconstructing how a system works.

Thematic fit, not the original investment rationale.

What matters in this area

Context freshness, traceability, and fit within developer workflows.

Context for agentic software development, connecting code with the knowledge around it.

Thesis connection

Code alone does not explain every engineering decision. Bringing team knowledge into development can help agents make changes that fit the wider system.

Thematic fit, not the original investment rationale.

What matters in this area

Context relevance, permission boundaries, and the quality of resulting changes.

Document-processing infrastructure for AI applications and agents.

Thesis connection

Documents sit inside many important business workflows. Turning them into reliable structured inputs can make automation useful beyond a demonstration.

Thematic fit, not the original investment rationale.

What matters in this area

Extraction quality, exception handling, and integration into downstream workflows.

A context layer connecting enterprise information and agent memory.

Thesis connection

Long-running agents need relevant history, not just a larger prompt. Context infrastructure can help preserve relationships and retrieve the information a task actually needs.

Thematic fit, not the original investment rationale.

What matters in this area

Memory relevance, temporal accuracy, and controlled access to enterprise context.

AI assistants that combine persistent context and connected tools to carry out work.

Thesis connection

Agentic software becomes valuable when it completes useful workflows with the right context and permissions. The opportunity is in execution, not conversation alone.

Thematic fit, not the original investment rationale.

What matters in this area

Task completion, user control, and continuity across recurring work.

Simulation and evaluation systems for learning and assessing agent behavior.

Thesis connection

An agent’s usefulness depends on how it behaves in realistic environments. Better simulation, evaluation, and feedback connect research progress with dependable execution.

Thematic fit, not the original investment rationale.

What matters in this area

Task representativeness, feedback quality, and behavior beyond a benchmark.

Evaluation, observability, and governance for AI systems.

Thesis connection

Deploying AI creates an ongoing measurement problem. Teams need to understand changes in quality and behavior as models, data, and workloads evolve.

Thematic fit, not the original investment rationale.

What matters in this area

Evaluation coverage, production visibility, and actionable quality signals.

Now Datadog Experiments

Experimentation infrastructure for measuring product and business impact.

Thesis connection

Shipping an AI feature is not evidence that it works. Experimentation connects technical changes to user behavior and measurable product outcomes.

Thematic fit, not the original investment rationale.

What matters in this area

Measurement integrity, experiment design, and relevance to business decisions.

AI-powered data workflows for modeling, analysis, and communicating insights.

Thesis connection

Analytical work needs business context and a sound method, not just generated answers. AI can become more useful when those practices are built into the workflow.

Thematic fit, not the original investment rationale.

What matters in this area

Analytical grounding, repeatability, and the path from question to decision.

Systems that synthesize data, environments, and feedback for AI training.

Thesis connection

As training becomes more automated, the quality of its inputs and feedback matters more. Systems that improve that loop can address a bottleneck beyond raw compute.

Thematic fit, not the original investment rationale.

What matters in this area

Training signal quality, useful task coverage, and generalization.

AI-native software development around a shared, executable model of intent.

Thesis connection

Generating code is only one part of engineering. A durable representation of intent can help teams understand, change, and govern what AI builds.

Thematic fit, not the original investment rationale.

What matters in this area

Faithfulness to intent, maintainability, and human control over changes.

A control and enablement layer for enterprise agents and their tools.

Thesis connection

Agents expand the range of actions software can take. Enterprise adoption depends on understanding those connections and governing access without stopping useful work.

Thematic fit, not the original investment rationale.

What matters in this area

Identity, permissions, runtime policy, and visibility across tool use.

AI agents for enterprise revenue workflows.

Thesis connection

Applied AI has to fit the work of a specific team. Revenue workflows offer a concrete setting to assess whether context and automation translate into useful action.

Thematic fit, not the original investment rationale.

What matters in this area

Workflow integration, information quality, and measurable operating value.

Security tooling for identifying and addressing risks in agentic AI systems.

Thesis connection

Agentic systems introduce risks across models, tools, and actions. Security needs to follow that lifecycle rather than treating the model as an isolated component.

Thematic fit, not the original investment rationale.

What matters in this area

Attack-surface coverage, prioritization, and practical remediation.