Frequently asked questions
A few things
worth knowing.
About the work, the technology and how we collaborate.
What does Knight CTO build?
AI systems that perform multi-step work: agent harnesses, autonomous swarms, self-healing software, voice and avatar agents, messaging agents, retrieval and memory systems, evaluation tooling, inference and cloud infrastructure. Engagements also cover product architecture, integrations and engineering leadership.
Will I work directly with Boris?
Yes. Boris leads the technical work and stays involved in architecture and implementation. The engagement defines responsibilities, deliverables and how he will collaborate with your existing team.
Do you only work in legal technology?
No. Legal AI is one application domain. The core work spans autonomous systems, software repair, agent infrastructure, model serving, cloud platforms and product delivery.
What has the autonomous swarm demonstrated?
The documented bounded proof completed an autonomous repair of a disposable application, followed by a rebuild, independent browser acceptance retest and promotion. Broader continuous operation and arbitrary application reliability remain separate goals; this is engineering work, not a generally available autonomous repair product.
How do you choose models and infrastructure?
Start with the workflow, data boundaries, evaluation criteria and operational requirements. The capability set includes LiteLLM routing, vLLM serving, local inference, AWS and Google Cloud. A project uses the components its requirements justify.
Can you work with an existing engineering team?
Yes. Work can be scoped around a difficult system, an AI capability, architecture and delivery support, or ongoing technical leadership. Ownership, handoffs and acceptance criteria are agreed with the team.
How are engagements priced?
Scope, deliverables, acceptance criteria and commercial terms are agreed before implementation. A first conversation establishes the problem and the right shape of engagement; there is no universal price for an unspecified system.
How do we get started?
Book a 15-minute conversation, use the longer 30-minute intake, or send a short description of the system, the users and the outcome you want. Share the problem first; confidential material can wait until the appropriate arrangements are in place.