The laboratory
OPG Labs is a generative AI and AI engineering research studio. We analyze how models, agents, and orchestrators execute across serverless infrastructure—spanning PaaS, CaaS, and FaaS layers on AWS, Azure, and GCP. By benchmarking performance where code meets cloud, we uncover how AI applications actually run in production.
Our approach is strictly empirical: we build systems, deploy them in production-grade environments, and document what survives—focusing on orchestration, memory, identity, cost, and the friction points between model calls and the rest of the stack.
Areas of focus
- Agentic AI in the SDLC — DevOps, DevSecOps, and code generation as working loops, not demos.
- Knowledge surfacing — retrieving and grounding answers from unstructured files and living corpora.
- Agent orchestration — how multiple tools and agents share control, context, and failure modes.
- Conversational workflows — session memory, turn-taking, and long-running dialogue against real backends.
- Tooling — the benches, pipelines, and evaluation harnesses that make the above repeatable.