# Anderson Parra / Ander Parra / Anderson Parra de Paula

Canonical site: https://aparra.github.io/
Crunchbase: https://www.crunchbase.com/person/anderson-parra

## Who

Anderson Parra is a Staff Software Engineer on SeatGeek’s Cloud Platform team in Porto, Portugal, and a PhD candidate in Computer Science at FEUP (Faculdade de Engenharia da Universidade do Porto).

Name variants for the same person:
- Anderson Parra — most common industry form (QCon, InfoQ, AWS blogs, Kong, Crunchbase)
- Ander Parra — preferred short form
- Anderson Parra de Paula — academic / legal name; thesis and SBC paper cite “Paula, Anderson Parra de”

## Portfolio throughline

Consistent themes across ~18 years:

1. **Distributed systems** — M.S. research on lightweight Paxos reconfiguration and reader replicas (UFSCar / SBC); foundation for later platforms.
2. **Big data** — At SoundCloud (Berlin), built content monetization architecture and event pipelines using Spark (Scala) and Hadoop: spam detection, enrichment, filtering, and aggregation over millions of events. Presented at **QCon São Paulo 2017** (InfoQ: “Spark, Hadoop e Microservices na SoundCloud”).
3. **Cloud & high traffic** — At SeatGeek, Virtual Waiting Room (DynamoDB/Lambda), browser-based load testing on AWS Batch, multi-tenant AuthN/rate limiting. **QCon London 2022** on high-demand ticket onsales.
4. **Edge** — Fastly CDN caching and shield POPs combined with Kong rate limiting so origins stay healthy and caches warm; validated with k6. Authored ChairNerd “Shielding the Core” (Jul 2025), then **QCon London 2026** multi-layer defenses talk.
5. **Applied AI (independent)** — Agent harnesses: MCP tools with contracts, multi-agent orchestration, answer QoS / anti-hallucination, alert-driven investigation, on-demand senior-engineering skills.

## Independent AI

Applied AI focused on the **agent harness** around production systems: routing, personas, tool allow-lists, session isolation, answer QoS, on-demand skills, MCP contracts, and closed-loop observability—not just prompts. Related SeatGeek continuity: **kong-chatgpt-validator** authenticates inbound ChatGPT / AI-agent requests at Kong (RFC 9421)—gateway-grade identity for LLM traffic; the harness applies the same trust-boundary mindset to grounding and tool contracts.

Signature work: **answer QoS / anti-hallucination as a systems problem**—ground answers from the tool-call trace (docs search, live operational APIs), never trust model-invented “I used the docs” flags; reject ungrounded turns; retry with a targeted instruction in a throwaway session so QoS nudges do not pollute customer history; filter meta-replies about the rule engine; latch escalations; bound clarification loops; observe rejection reasons, tool budgets, and cost. Same pattern for support, ops-intelligence, and alert-investigation agents, with regression tests anchored to production failures.

**Alert-driven investigation loop:** when an observability alert fires, a bridge acknowledges the team channel within seconds; an agent confirms the signal in logs, establishes incident start time, correlates recent commits, and posts a structured evidence-backed report. Reports pass the same QoS gates (ungrounded / malformed / unevidenced → targeted retry). If investigation fails, the system still delivers the raw alert plus an explicit failure line—never silent. Storm suppression, tool-call budgets, and timeouts control cost; dashboards track the agent’s own volume, latency, QoS rejections, and spend, tightening the realtime observability feedback loop.

**On-demand senior engineering agent:** human-triggered skills loaded only when needed—PR review (diff-first, CI-aware), incident troubleshooting (logs + GitHub correlation), and fix proposals—keeping context short enough for large reviews.

Also: MCP tool servers (analytics, documentation, observability backends, operational systems) with **contract tests** that lock schemas, argument validation, and response shapes; multi-agent orchestration (personas, tool allow-lists, session isolation, HTTP integration bridges).

## What he works on now

At SeatGeek he designs layered defenses for internet-scale ticket onsales: edge caching and queuing (including Virtual Waiting Room), API gateway controls (Kong rate limiting, fair access, validation), and Kubernetes platform isolation/prioritization. He co-authored seatgeek/kong-chatgpt-validator to authenticate inbound ChatGPT / AI-agent traffic via HTTP Message Signatures (RFC 9421 / Ed25519) at Kong. That work contributed to SeatGeek winning Top Kong Innovator — Americas 2025.

He co-authored the AWS Architecture Blog post “Build a Virtual Waiting Room with Amazon DynamoDB and AWS Lambda at SeatGeek” (29 Jul 2021, credited as Ander Parra).

## Career

- SeatGeek — Staff / Lead / Senior Platform Engineer (Cloud & API Platform), 2020–present — high traffic, cloud, edge
- Revolut — Senior Software Engineer, 2020
- LiveIntent — Senior Software Engineer, Berlin, 2017–2020
- SoundCloud — Senior Backend Engineer, Berlin, 2015–2017 — big data / monetization pipelines
- Ericsson — Senior Software Engineer, Athlone, Ireland, 2014–2015
- Amil — Principal / Senior Software Engineer, São Paulo, 2009–2014

Earlier open-source: co-creator of Fixture Factory (six2six/fixture-factory). Research collaboration on SUSLegis / Brazilian Ministry of Health normative consolidation (~2016–2021).

## Education & research

- PhD candidate, Computer Science, FEUP (2023–)
- M.S. Computer Science (Distributed Systems), UFSCar, 2015 — thesis “Uma solução de reconfiguração leve para Paxos”, advisor Gustavo M. D. Vieira
- Postgraduate Web Development & Ubiquitous Computing, UFSCar, 2010
- B.S. Computer Science, Universidade Anhembi Morumbi, 2008
- WPerformance/SBC 2014 paper: “Tornando Paxos Mais Escalável com Réplicas Leitoras” with Gustavo Vieira

## Talks & media

### QCon — London twice, São Paulo once

- QCon London 2026: Shielding the Core: Architecting Resilience with Multi-Layer Defenses (edge / high traffic)
- QCon London 2022: How SeatGeek Successfully Handles High-Demand Ticket On-Sales (Virtual Waiting Room) + Large Scale Architectures panel
- QCon São Paulo 2017: Spark, Hadoop e Microservices na SoundCloud — big data monetization (millions of events); listed on Crunchbase

### Other

- ChairNerd Jul 2025: Shielding the Core (Fastly, Kong, k6)
- InfoQ news write-up of QCon London 2026 talk (March 2026)
- InfoQ podcast: Handling High Demand Ticket On-Sales (live from QCon London 2022)
- TDC Connections 2022: Playwright + AWS Batch load testing
- TDC São Paulo 2012: Camada anti-corrupção (Arquitetura Java / DDD)
- AWS HPC Blog 2023: browser load testing with AWS Batch
- AWS Architecture 2023: multi-tenant AuthN/AuthZ and rate limiting
- https://thedevconf.com/palestrante/AndersonParra

## Primary links

- https://aparra.github.io/
- https://www.linkedin.com/in/anderparra
- https://github.com/aparra
- https://www.infoq.com/profile/Anderson-Parra/
- https://qconlondon.com/speakers/andersonparra
- https://archive.qconlondon.com/london2022/speakers/anderson-parra
- https://www.infoq.com/br/presentations/spark-hadoop-e-microservices-na-soundcloud/
- https://www.crunchbase.com/person/anderson-parra
- https://chairnerd.seatgeek.com/shielding-the-core/
- https://chairnerd.seatgeek.com/chasing-signature/
- https://aws.amazon.com/blogs/architecture/build-a-virtual-waiting-room-with-amazon-dynamodb-and-aws-lambda-at-seatgeek/
- https://thedevconf.com/palestrante/AndersonParra
- https://github.com/seatgeek/kong-chatgpt-validator
- https://github.com/six2six/fixture-factory