loop/01 — engineering identity vietnam — singapore

Gia Huy Du

Scale the systems. Rewire the engineering.

Agentic Backend Engineer — 7+ years of Go, Ruby, and distributed systems serving 100M+ users at 200k+ RPS. Building AI-native, spec-driven engineering. Singapore.

~/core/runtime live
peak rps
200k+
users
100M+
throughput
40x
stack
go · ruby
  • gw p99 routing 42ms
  • redis pipeline p50 3ms
  • etl events/s 38.9k/s
  • search ctr 68.4%
  • k8s pods live 221
  • go/gc pause p99 236μs
  • grpc rps 41.8k/s
  • cdn cache hit 92.5%
loop/02 — what i do

Built scalable systems. Now rewrire them with AI.

Expertise over 7+ years in backend engineering.

  1. 01

    Build the platform.

    I design and ship the APIs that power the platform's key features — the endpoints users hit every day.

    • go
    • ruby
    • domain-driven design
    • system-design
  2. 02

    Go below the API.

    When things are slow, I dig past the framework to fix it at the source — tuning the garbage collector, writing custom Redis commands in C, and rewriting the slowest paths in the codebase. The real wins are usually hidden underneath.

    • redis module-capi
    • low-level optimization
    • profiling
    • load-testing
  3. 03

    Reliable at scale

    When things break, I'm the one on the call — finding the root cause, getting it back up fast, and making sure it doesn't happen twice.

    • incident-response
    • on-call runbook
    • k8s
    • datadog
    • observability
  4. 04

    AI-native Engineering

    I make AI a real part of how engineers ship, not a demo. Spec-driven development by default, agent workflows tuned to the codebase, and workshops that turn curiosity into shipped features.

    • claude code
    • github copilot
    • spec-driven development
    • agentic-workflows
loop/03 — proof

Achievements, in numbers.

  1. 0 k+
    peak rps

    What full-scale personalization demands, proven in production at Rakuten Viki. Sustained with no bottleneck at this level — and room to go further.

  2. 0 M+
    users

    Viewers spread across 200+ countries, carried by the Content domain I led through every peak.

  3. 0 x
    throughput

    Rewrote Edge API from Node.js to Go: 264 pods became 22, throughput jumped 40x.

client api gateway edge clusters × 11 regions central cluster origin data storage web mobile tv apps api gateway content search personalize ratings redis distributor indexer rabbitmq etl postgres
ingress service mesh state pipeline core services · simplified
loop/04 — case study

The Edge Compute rewrite, in three states.

state/01 — the problem

264 pods, one hot path.

fleet
264 node.js pods
scale
reached limit at 5k RPS
cost
too high for scale up
state/02 — the decision

Rewrite the edge in Go.

garbage collector
tuned
data model
redis pipelined, denormalization
golang internal
optimize json deserialization
state/03 — the result

40x throughput. Same hardware.

fleet
22 go pods
throughput
40x
peak
200k+ rps
loop/05 — track record

Career so far.

Three roles, same pattern — every system I touched left in better shape than it started.

  1. 2023—2026 Rakuten Viki Senior Software Engineer

    Content domain: gateway, distribution, ETL pipeline. 100M+ users, 200k+ RPS. Top Innovation Engineer 2024.
  2. 2020—2023 Ascenda Senior Software Engineer

    Loyalty orchestration layer integration for 8 financial institutions — $10M+ annual B2B revenue. Top Performance Engineer 2022.
  3. 2019—2020 Dwarves Foundation DevOps Engineer

    Cloud Product API in Go. Kafka→NATS migration.
side quests
  • copy trading engine

    Go, 10–20ms order replication over Binance WebSocket. 54M USDT volume in 30 days.

  • mmtk ruby bindings →

    GC internals in C and Rust — ported to macOS, made Ractor-safe.

rubyconf taiwan + thailand '23 · gophercon vietnam organizer '19 · mit iquhack '24 '25

loop/06 — writing

Field notes.

all notes →

Writing is the eval suite. Real systems, real numbers, posted as they happen.

loop/07 — contact

Say hello.
Bring me hard problems.

Open to senior / staff roles on high-scale backend platforms. Available for consulting on Go, Ruby, distributed systems, and AI-native engineering practice.