Distributed systems for a graph database

Overview

At Aerospike I worked across the graph database's backend and developer tooling: stabilizing query distribution under load, cutting latency in a cloud-hosted fraud detection service, and building infrastructure that let LLM agents talk to the graph directly through a Model Context Protocol server.

github.com/aerospike/aerospike-graph →

Highlights

• 01

  Implemented distributed load balancing algorithms and unit tests across three Gremlin drivers, improving system reliability under high concurrency by 60%.

• 02

  Debugged and refactored the cloud-hosted fraud detection service, integrating multithreading and decoupling patterns that cut backend latency by 70% and increased throughput 3.4×.

• 03

  Built and containerized an MCP server enabling LLM/agent integration with a graph database, presented to project managers and stakeholders.

• 04

  Built a dynamic, multithreaded data generator producing synthetic customer data at over 1M rows/sec, accelerating benchmarking and customer POCs by 2.6×.

• 05

  Built a CI/CD pipeline in GitHub Actions for automated unit testing of 12+ applications, reducing regression errors by 40%.