I'm Sajjad, co-founder and CTO of DaaX. Excited (and a little nervous) to share LAKEer with you today. What inspired us to build this? This has been brewing for over a decade. As Global Practice Head for Cloud, Data & AI at Wipro, I was consumed by augmenting neural networks with external knowledge representation using knowledge graphs — infusing domain understanding to fight off hallucinations and inject explainability and causal reasoning. We called it the Enterprise Domain Brain, and we discovered applications everywhere: supply chain, marketing, chemical & material science knowledge discovery, and more. Then Generative AI arrived, and we watched hundreds of billions of dollars pour into it. It became clear to us that this investment needed a counterweight: Semantic Verifiers — the missing layer that enables governed use of Generative AI in the enterprise. That conviction became LAKEer. What problems were we trying to solve? Hallucination isn't a prompt problem, it's an architecture problem. Retrieval-augmented generation alone doesn't fix it. So we built ClaimGuard, a multi-stage verification pipeline that checks every claim in an answer against grounded sources before it reaches the user. Enterprise knowledge isn't just text. It lives in tables, entities, relationships, and domain semantics. LAKEer combines knowledge graphs with six-channel retrieval — vector, keyword, graph, tables, entities, semantic units — so answers draw on structure, not just similarity. Generic AI doesn't speak your domain. LAKEer supports pluggable domain ontologies and business rules that enhance information extraction, query understanding, and answer verification — so the system understands your industry's concepts, not just its vocabulary. On the FACTS Grounding public benchmark (860 examples), LAKEer scored 77.7% accuracy with 92% eligibility — as a complete system, not a bare LLM. How did our approach evolve? We started thinking of this as "better RAG." Over time we realized the real product isn't retrieval — it's reliable information supply. As AI agents proliferate, they all need a trustworthy substrate to draw from. So LAKEer evolved into that layer: agents (via our MCP interface — 16 tools out of the box) consume verification-gated answers with full provenance instead of raw search results. We also went deep on the neuro-symbolic side — because we found that symbolic structure is what makes verification tractable, not a nice-to-have. We'd love your feedback on: where verifiable AI matters most in your work, and what would make you trust (or distrust) an AI answer in production. Ask me anything — I'll be here all day! 🙏

DaaX/LAKEer

The trusted knowledge layer for AI agents.

The trusted knowledge layer for AI agents.