Applied AI research for structured semantic reasoning

We build foundational technology that gives AI true structural understanding of complex information.

Founded by Andres Martinez. 15+ years at Microsoft, Principal Architect on the VP/CTO team. Multiple patents in enterprise AI systems.

Research Thesis

The reasoning gap

Current AI systems retrieve fragments and generate answers from whatever they find. The language model has no way to know if the retrieval missed something critical. It reasons over incomplete information without knowing it's incomplete.

This works for simple questions. It breaks on complex analysis: legal cases spanning hundreds of documents, research synthesis across dozens of papers, enterprise decisions that depend on connecting information across organizational boundaries. In these domains, incomplete retrieval produces incomplete reasoning. And nothing in the system tells you what's missing.

We're closing this gap. Kacti AI builds the reasoning layer between retrieval and generation, so AI systems can reason over structure, trace relationships, detect gaps, and report what they found and what they couldn't find.

Our Approach

Three convictions that guide our research

Structure first, then reason

Information must be structured semantically before reasoning can be reliable. We build connected models of complex information, not embedding indexes, so the reasoning layer works with real relationships, not statistical similarity.

Retrieval is a reasoning task

In current systems, retrieval and reasoning are separate phases. The retriever finds content by pattern matching; the language model reasons over whatever it receives. We unify these phases so retrieval itself becomes a reasoning process, guided by query intent and evaluated for completeness.

Show what's missing, not just what's found

Every analysis should report its own limitations. Our systems tell you what evidence was found, what couldn't be found, and where contradictions exist. Confidence without transparency is a liability in high-stakes domains.

Structure before code. Meaning before implementation. We model a domain's essential concepts and relationships first, then strip everything else away. What reaches the reasoning layer is signal, not noise. That discipline is why our systems find what others miss.

Our Current Focus

Legal Case Intelligence

We chose legal because it's a domain where our technology can deliver significant value. A missed precedent, an undetected contradiction, a gap in the evidence chain. These aren't inconveniences. They're case outcomes.

We work with litigation teams on active cases, structuring facts, mapping claims to legal standards, and stress-testing arguments before opposing counsel does.

Looking Ahead

Foundational technology

The reasoning systems we're building are domain-agnostic at the engine level. Legal is first because it demands the most from our technology: complex multi-document reasoning, structured argumentation, gap detection across large evidence sets.

The same foundational systems will extend to other domains where structured semantic reasoning creates value: enterprise operations, research synthesis, compliance analysis. We're building the foundation first, then expanding the surface area.

Talk to the Founder

We work with a small group of litigation teams. If you're working in a domain where incomplete reasoning has real consequences, let's talk.