Software Engineer · Data Infrastructure

Roman Garza

Kafka event pipelines, cloud migrations, and backend services

Software engineer with 7 years of experience building backend services and data infrastructure. At Indeed I worked on company-wide Kafka event infrastructure used by dozens of teams, led a migration of event-dependent services to AWS, and designed and shipped a service that downstream data jobs depend on. Before that, at Quantea, I built DPDK-based packet capture products as a lead engineer.

Select an employer to highlight its work.
over $1.4 millionin appliance sales from 2 new products I led backend development for at Quantea
10 timesfaster packet capture, from 1–2 Gbps to 15–20 Gbps
about $20,000a year in query costs saved by a new Presto/Trino parser (about $1,700/month)
29%of event-record writes eliminated by turning off 2,843 event types (about 16% of stored bytes)

Experience

Sep – Oct 2026
Saint Paul's Monastery Icon Shop · Volunteer e-commerce engineer, one-week engagement
Oct 2021 – Aug 2025
Indeed · Software Engineer II, Data Infrastructure
Dec 2018 – Oct 2021
Quantea · Lead Software Engineer, packet capture appliances

Selected work

Internal details are generalized. Diagrams are simplified sketches of how data moved, not architecture docs.

Partition Ready Events

Indeed · 2023 · design and delivery

Downstream jobs such as index builders needed to know when a partition of event data was complete and safe to consume from the internal storage layer. I designed and shipped a Kafka service that announces exactly that.

  1. Event storage layer
  2. Partition Ready Events
  3. Readiness events on Kafka
  4. Downstream jobs (e.g. index builders)
Simplified flow
  • Gathered requirements from 4 consuming teams, wrote the design docs, and presented them to the team and stakeholders for feedback.
  • Evaluated data-structure designs for processing log-volume metrics events with a teammate and chose the approach the service shipped with.
  • Took the MVP to production, then reproduced and profiled the service to fix processing bottlenecks, missing events, and offset-seeking bugs.
  • Led its operational readiness review and designed changes to guarantee event delivery.
KafkaEvent Metadata Service

Moving event metadata services to AWS

Indeed · 2022 · led

Three event-dependent services ran on on-prem MySQL, and moving them to AWS was a prerequisite for decommissioning the legacy data center. I led the migration, wrote the project design, and broke the work down across engineers.

  1. Services on on-prem MySQL
  2. Aurora databases created (QA and prod replicas)
  3. Production cutover validated
  4. Legacy data center decommissioned
Delivered incrementally, with results validated at each step
  • Created the MySQL Aurora databases and coordinated QA and production replicas with the SQL and platform teams.
  • Migrated Typecode Service, Storage Location Service, and EventViewer, and validated the production database cutover.

Moving all three services to Aurora cleared them out of the legacy data center ahead of its decommission.

AWS AuroraMySQLTerraform

Kafka Exchange UI

Indeed · led

I designed and led a subscription-based web app that simplified company-wide onboarding to the Kafka event pipeline.

  1. Engineering teams
  2. Kafka Exchange UI
  3. Event pipeline subscriptions
Simplified flow
  • Owned the data model, UI/UX collaboration, ticket breakdown, infrastructure as code, and code review.
  • Mentored an intern through implementation, took the app to production, and presented it to the wider Data Infrastructure organization.
Node.jsReactTypeScript

Tracking down duplicate event types

Indeed · 2023 · incident

Duplicate event types started appearing in the Event Metadata Service. I analyzed the incident independently and put correcting the data ahead of new feature work.

  1. Duplicate types found
  2. Impact analyzed
  3. Race condition fixed with a Postgres sequence
  4. Data correction prioritized, retrospective written
Sequence of the response
  • Traced the duplicates to a race condition in type-code creation and fixed it with a Postgres sequence generator.
  • Wrote the incident retrospective.
  • Delivered the service's production readiness: full legacy API support, API key authentication, and the cutover plan.
JavaPostgreSQLS3

Cutting data volume and query cost

Indeed

Two changes that reduced what the pipeline stored and what its queries cost.

  1. Benchmark plan with exit criteria
  2. Correctness tests with the Data Lake team
  3. New default Presto/Trino parser
  4. ~$1,700/month saved
Parser rollout
  • Built bulk event-type updates that let the team turn off writes for 2,843 event types, which made up 29% of records and about 16% of bytes in Q4 2022.
  • Built a benchmarking plan with exit criteria for a new log parser, ran correctness tests with the Data Lake team, and kept stakeholders updated weekly before making it the default.

Query costs dropped by about $1,700 a month.

Presto/TrinoKafka

Packet capture appliances

Quantea · 2018–2021 · lead engineer

I designed and implemented applications for packet capture, dissection, and storage, and led DPDK-based Network Packet Broker development as subject matter expert for 5 stakeholders.

  1. Capture
  2. Dissection
  3. Storage
The pipeline each appliance ran
  • Raised packet capture throughput from 1–2 Gbps to 15–20 Gbps (10×) with DPDK, which helped win long-term contracts, and expanded filtering options 4×.
  • Designed the capture, dissection, and storage applications behind 2 new appliance products, which generated $1.4M+ in sales.
  • Cut Network Packet Broker development time by 25% as DPDK subject matter expert for 5 stakeholders.
C/C++DPDKLinux

Fixing a monastery's web store

Saint Paul's Monastery · Boscobel, WI · Sep–Oct 2026 · volunteer, one week

The Icon Shop of Saint Paul's Monastery runs on WordPress and WooCommerce. Over one week I cleaned up its catalog, fixed problems customers saw on the live store, and left the staff a guide so they can run it themselves. I used Claude to speed up investigation and implementation, and verified each change on the live store before rolling it out.

  1. Inventory shown wrong on the live store
  2. WooCommerce hooks traced
  3. Custom snippet found
  4. Fix confirmed past caching
The storefront bug, from symptom to cause
  • Reconciled inventory across 1,000+ SKU and variation records, found missing catalog entries, and built CSV bulk-import workflows to create and update variations (SKU, attributes, pricing, stock).
  • Debugged failed variation imports across parent/child products, global attributes, variation metadata, visibility, and stock settings, testing on a small set of products before updating the whole catalog.
  • Traced a live storefront inventory bug through WooCommerce hooks to a custom snippet on the woocommerce_get_availability filter, and confirmed the fix past WordPress caching.
  • Fixed the store's inability to mark customers as tax-exempt by setting up a WooCommerce tax-exemption plugin.
  • Wrote a plain-English help guide and training material for the non-technical staff who manage the store.
WordPressWooCommerceCSV importClaude

Also at Indeed

What I work on

Distributed systems

Services other systems rely on to be correct, like knowing when a partition of event data is complete.

Partition Ready Events · Duplicate types

Data infrastructure

Kafka event pipelines, storage and query systems used by dozens of teams.

Systems map · Data volume

Performance

Finding bottlenecks and cutting cost, from profiling services to a cheaper query parser and 10× faster capture.

Query cost · Packet capture

Production engineering

Migrations, monitoring, incidents and readiness reviews: the move to AWS, SLO monitors for 6 services, retrospectives.

AWS migration · Incident response

Systems map

How the systems behind my case studies connect. Arrows follow the data. Select a box to see my part in it, then open the write-up.

How to read this

Hover, tap or tab to a system

Each box is a system from the case studies below. Its connections light up, and you can jump to the write-up from here.

  • Built or led
  • Worked on
  • Context, not mine
Read the map as a list
  • Indeed: event pipeline
    • Engineering teams → Kafka event infrastructure
    • Engineering teams → Kafka Exchange UI
    • Kafka Exchange UI → Kafka event infrastructure
    • Platform tooling → Kafka event infrastructure
    • Kafka event infrastructure → Event storage
    • Event storage → Partition Ready Events
    • Partition Ready Events → Downstream jobs
    • Event storage → Presto/Trino
    • Event Metadata Service → Partition Ready Events
  • Indeed: move to AWS
    • 3 metadata services → AWS Aurora
  • Quantea: packet capture appliances
    • Capture → Dissection
    • Dissection → Storage

Skills

Languages
Java, C/C++, Python, JavaScript/TypeScript, Bash
Data & streaming
Kafka, Schema Registry, Presto/Trino, MySQL, PostgreSQL, AWS Aurora
Cloud & infrastructure
AWS, Terraform, Kubernetes, Linux, DPDK
Frameworks & tooling
Spring Boot, Node.js, React, Gradle, GitLab CI, Datadog, Git, WordPress/WooCommerce
AI-assisted work
Claude, for investigation and implementation

Education

2018
B.S. Computer Science · Oregon State University
2015
B.S. Biology · Missouri State University