Projects that prove the depth.

Problem

TRC files, AWR reports, and stack traces are noisy. Triage often starts with scattered clues across database, JVM, and network layers.

Built

A Spring Boot service that parses raw diagnostic artifacts, classifies failures by layer, and emits Markdown plus JSON root-cause reports for downstream tooling.

Impact

Turns production failure data into a repeatable diagnostic workflow that can plug into support or CI/CD systems.

Problem

Database behavior changes under sustained concurrency. Manual scripts rarely capture TPS, p95/p99 latency, pool pressure, and failure classes together.

Built

A configurable Spring Boot framework using ExecutorService pools, YAML load profiles, WebSocket metrics, and post-run failure classification.

Impact

Productizes the kind of workload thinking I used at Oracle while stress-testing RDBMS behavior at billion-row scale.

Problem

Vector search decisions are often made from vibes: index choice, recall, latency, and memory trade-offs need reproducible measurement.

Built

A Spring Boot benchmark suite comparing HNSW, IVFFlat, and exact search across dataset sizes, dimensions, and query profiles.

Impact

Brings Oracle vector-search benchmarking discipline into a public pgvector project with readable CSV and Markdown outputs.