Data & AI Expert at SE-Link, where the focus is making AI dependable enough to act inside real engineering projects - plant lines, infrastructure, and industrial automation. The throughline of the work is neurosymbolic: pairing language models with sound symbolic verification on top of Neo4j knowledge graphs, so autonomous actions are bounded by proof and calibrated trust rather than probabilistic confidence.

This includes deterministic, provenance-bearing GraphRAG, bitemporal graph memory, and solver-gated multi-agent systems deployed against messy, multi-modal engineering data. Kani cares about the gap between impressive demos and systems you can actually put near a factory floor, and enjoys sharing the patterns and the hard-won mistakes - that close it