Independent research: GraphRAG makes AI agents 80% more truthful | Read the report

NODES 26 — November 12, 2026

Merging Under Load: Point-in-Time Entity Resolution on a Live Neo4j Graph

Session track: Data Intelligence

Session time:

Session description:

Entity resolution looks like data cleaning: cluster the duplicates, elect a canonical, merge, keep an audit trail. This usually works - until the graph is read every second by AML rules, sanctions screening, and analysts on open cases. Thus it's not data cleaning, it's a concurrency problem. And Neo4j doesn't give you point-in-time consistency for free. You build it yourself, in Cypher. Sunidhi grounds this in TerraLens, a transaction monitoring and screening engine for a large regulated financial services provider: nearly 60 million transaction relationships, growing by the second, re-resolved every hour while live services read and write it. Every obvious escape is closed. You can't pause the graph. You can't rebuild from scratch, it takes hours and erases merge history regulators can demand. You can't destructively merge an entity someone may be mid-investigation on. Something is always mid-read, so the merge has to land under live queries without them noticing. What you'll take away: treat a merge as a concurrent mutation, not a batch edit; hold a consistent point-in-time view for a reader while the entity it's reading is merged away; make the merge status-aware, so a node under investigation is handled differently from a settled one; mutate a heavily-read graph on a schedule while keeping a durable, queryable audit trail. Sunidhi will be candid about what the design leaves to human review, and where it stops.

Speaker

photo of Sunidhi Shende

Sunidhi Shende

Software Engineer, RUDRA

Sunidhi Shende is a software engineer at RUDRA. She is a systems builder with interests spanning AI, cyber-security, distributed systems and infrastructure at scale. In the past couple of years, she has worked extensively with the maritime industry developing a cyber-security and network management platform, and with the financial industry dealing with transaction screening and monitoring of over 60 million transactions. She is passionate about building resilient, scalable systems that solve complex real-world challenges and create meaningful impact. Outside of engineering, she enjoys adventurous travel and connecting with people and ideas from diverse tech communities.