Zeba Ali · Noida, India · open to backend roles

Backend engineer working at the database and AI-infrastructure layer.

Four years inside Oracle's RDBMS engine — query processing, execution plans, and the first parameterized vector-search benchmarking engine at Oracle, measuring recall and latency across 500+ HNSW/IVF configurations at billion-row scale.

zebaali1415@gmail.com GitHub LinkedIn

APIrequestSERVICElogicEVENTasyncDATApersistCACHEfast path

Oracle RDBMS4 yearskernel engineering

StatusBuildingin public

What I work on

Database internals

Oracle RDBMS source code — SQL parser, query transformation, execution plan analysis, access-path and index selection, MVCC, B-tree and hash indexing.

SQL parser
Query transformation
Execution plans
Access-path selection
MVCC
B-tree · hash indexing

Vector search & AI infrastructure

HNSW and IVF index behavior, recall@k versus latency tradeoffs, workload characterization at billion-row scale, pgvector.

HNSW · IVF
recall@k vs latency
Workload characterization
Billion-row scale
pgvector

Distributed backend systems

Java 17, Spring Boot, Kafka event-driven architecture, Redis, Kubernetes, and Prometheus/Grafana/OpenTelemetry observability.

Java 17 · Spring Boot
Kafka
Redis
Kubernetes
Prometheus · Grafana
OpenTelemetry

I work on the layer most engineers never touch: how databases actually execute queries, and how vector indexes trade recall for latency at scale.

Projects

Database Stress Testing Framework

Database performance regressions are hard to catch because you cannot compare runs reliably.

Repository

Architecture — drawn from the repository.

Problem: Database performance regressions are hard to catch because you cannot compare runs reliably — different load, different data, no plan-level visibility.
Approach: A configurable JDBC load-testing engine with execution-plan capture per query, slow-query identification against configurable thresholds, index vs no-index comparison runs, and connection-pool exhaustion testing under varying concurrency.
Outcome: Repeatable run-over-run comparison with per-query p50/p95/p99 latency, TPS, and error rate — instrumented end to end with Prometheus metrics, Grafana dashboards, and OpenTelemetry tracing across query execution paths.

Java 17
Spring Boot
JDBC
Redis
WebSocket
Prometheus
Grafana
OpenTelemetry

OrderFlow

Distributed order flows fail in ways monoliths do not — partial failures, duplicate events, lost messages.

Repository

Architecture — drawn from the repository.

Problem: Distributed order flows fail in ways monoliths do not — partial failures, duplicate events, lost messages between database commit and message publish.
Approach: Six Spring Boot microservices with Saga-based order orchestration over Kafka using choreography, a transactional outbox to eliminate the dual-write problem, idempotent consumers with retry and DLQ, Redis GEO for driver matching, and WebSocket live tracking.
Outcome: Full order lifecycle verified end to end in ~1.3 seconds, including the failure path — a payment decline correctly compensating to CANCELLED. Deployed on Kubernetes with health probes and horizontal autoscaling, with Prometheus/Grafana/OpenTelemetry observability across all services.

Java 17
Spring Boot 3.2
Kafka
PostgreSQL
Redis
Kubernetes
Docker
Prometheus
Grafana
OpenTelemetry

Log Intelligence Tool

Incident triage means correlating signals that live in different places, manually, under time pressure.

Repository

Architecture — drawn from the repository.

Problem: Production incident triage means correlating signals that live in different places — application logs, database traces, kernel errors — and doing it manually under time pressure.
Approach: A REST service that parses database and kernel logs, performance reports, execution traces, and Java stack traces, correlates signals across application, database, and kernel layers, and classifies failures by layer.
Outcome: Structured root-cause reports across 7 failure layers, built directly from four years of Fortune 500 production incident analysis. Instrumented with Prometheus, Grafana, and OpenTelemetry across the analysis pipeline.

Java 17
Spring Boot
PostgreSQL
REST
Docker
Prometheus
Grafana
OpenTelemetry

Experience

Software Engineer II · Oracle India Pvt. Ltd.

Oracle Server Technologies Group — Oracle RDBMS, Noida

Jul 2022 — Apr 2026

Modified Oracle RDBMS source code to ship 4 production database features across the 19RU, 23RU, 23ai, and 26.1 releases — including parser-level support for nested CTEs, enabling MongoDB aggregation pipeline translation to Oracle SQL
Built Oracle’s first parameterized vector-search workload engine — 500+ HNSW/IVF configurations measuring recall@k, throughput, and p50/p95/p99 latency at billion-row scale. Adopted org-wide.
Engineered distributed validation infrastructure processing 5 billion rows across Oracle RAC, Exadata, and Exascale with 100 concurrent database instances, surfacing MVCC edge cases and deadlock patterns that fed Oracle 23ai hardening
Sole responder for 10+ Fortune 500 production incidents across banking, pharmaceutical, and industrial customers
Resolved 200+ defects and shipped 13 production fixes

Previously: Software Engineer Intern at Wabtec Corporation (2021—2022) — Java backend data pipelines and SQL Server integrations for rail-logistics workflows.

Immediate joiner · open to backend and database engineering roles

Let's talk about the systems layer.

zebaali1415@gmail.com

GitHub LinkedIn