Ronny Haryanto
AI-native Principal Engineer, Melbourne, Australia

[email protected]
github.com/ronny

Ronny Haryanto

AI-native Principal Engineer

I build and run production systems, and help teams build theirs.

Principal engineer with 20+ years building and running production systems. I set platform direction across teams and stay hands-on in Go, Kubernetes and infrastructure as code. I work AI-native: I ship LLM features to production and build with coding agents every day.

20+

years in production

60m

accounts migrated, no downtime

1.2b

events a day through one pipeline

3×

faster model calculations at UKG

Experience 2001–2026

  1. Independent project: AI-native small-business operations system

    In production

    AI-native small-business operations system

    • Every model call goes through OpenRouter, with the model chosen per request. One OpenTelemetry span per model call with GenAI semantic conventions. Every call, failures included, kept in a ledger with billed cost and tokens. Hybrid retrieval, async generation through Pub/Sub, and an MCP server with OAuth 2.0 and per-user consent.
    • Go
    • OpenTelemetry
    • Pub/Sub
    • MCP

  2. Independent project: Takeoff as a Service

    Pre-launch

    Takeoff as a Service

    • Reads construction plan PDFs and produces quantity takeoffs. Text pass first, vision-model fallback. Hand-checked answer sets with automatic scoring drive prompt and model changes (96% accuracy).
    • Vision models
    • Evaluation

  3. Such

    Principal Software Engineer

    Jun 2025 – Sep 2026

    Remote, Australia

    part of House of Doge

    Consumer neobanking app with Dogecoin integration. Team of 15 engineers, 14 Go services on AWS EKS and 10 shared libraries.

    • Platform and paved road. Set the backend architecture and wrote most of the shared platform, so engineers from any team could build and change any service the same way, faster and with fewer mistakes. APIs and messages are typed protobuf, shared by servers and clients and generated for almost any language. The platform includes:
      • a ConnectRPC API gateway that carries request context through to every service
      • the typed ConnectRPC client in the React Native app, wired into TanStack Query
      • a Go service template (ConnectRPC, OpenTelemetry)
      • shared libraries for servers, SQS consumers (dead-letter queues, backoff, FIFO) and context propagation
      • a one-command local stack for all services
      • an internal engineering portal behind Tailscale for deployments, ops dashboards, a service map, account management and debugging tools, with an audit trail and an MCP server
      • a reusable GitHub Action for CDK diff and deploy
    • Event platform. Designed and built doge-engine, the gateway between all services and the Dogecoin network. It runs a node fleet across AWS and other clouds, fans out transaction broadcasts, publishes an at-least-once event stream (EventBridge → dedupe and canonicalise → SNS/SQS) and feeds ClickHouse through Firehose and Parquet on S3 with ~5-minute latency.
    • Ledger. Designed the money movement on a TigerBeetle double-entry ledger. Staged inbox/outbox accounts and linked transfers make each state change atomic, including blockchain reorgs.

    14Go services on one paved road

    • Go
    • TypeScript
    • React
    • AWS
    • Kubernetes
    • Infrastructure as code
    • CI/CD
    • APIs
    • Event streaming
    • ClickHouse
    • TigerBeetle
    • User analytics
    • Observability
    • Payments & ledgers
    • Leadership

  4. Lyrebird Health

    Principal Software Engineer

    Jan 2025 – Jun 2025

    Melbourne, Australia (hybrid)

    AI medical scribe for clinicians.

    • Owned the backend architecture of an AI medical scribe (Node.js, TypeScript, PostgreSQL, AWS).
    • Integrated with Electronic Medical Record systems through HL7 and FHIR, with partners and customers.
    • Mentored other engineers, and did a lot of the design discussions and reviews.

    HL7 · FHIREMR integrations

    • TypeScript
    • PostgreSQL
    • AWS
    • Infrastructure as code
    • LLM systems
    • APIs
    • Leadership

  5. UKG

    Senior Principal Architect

    Feb 2023 – Jan 2025

    Remote, Australia

    Strategic Workforce Planning, formerly Quorbit

    Strategic workforce planning inside the UKG Workforce Management suite.

    • Led the work on performance and long-term scalability of model calculations. Cut calculation time from 1 hour to 20 minutes (Go, Kubernetes, GCP, gRPC, PostgreSQL).
    • Led a team of ~6 engineers. Delivered new features and integrated Strategic Workforce Planning into the UKG Workforce Management suite.

    60 → 20 minmodel calculation time

    • Go
    • Kubernetes
    • GCP
    • Infrastructure as code
    • APIs
    • PostgreSQL
    • Leadership

  6. GetSetUp

    Head of Platform

    Before that: Lead Software Engineer (2021)

    Apr 2021 – Feb 2023

    Remote, Australia

    Live online learning for older adults.

    • Set the platform architecture that moved the existing apps to Kubernetes on AWS. Wrote the patterns for observable, production-ready services.
    • Led ~8 engineers to launch the flagship web app on a mix of new and legacy services, then replaced the legacy services and internal apps with no stop in service (GraphQL, Next.js, Algolia, Aurora PostgreSQL, DynamoDB).
    • Built an events pipeline with strict schema validation, from client producers to self-serve analytics, at 1m events per day (Kinesis Firehose, Glue, S3, Athena, Metabase).
    • Built the entitlements system for retail and partner products on Stripe and DynamoDB.

    ~8engineers led to launch

    • TypeScript
    • React
    • AWS
    • Kubernetes
    • Infrastructure as code
    • APIs
    • PostgreSQL
    • DynamoDB
    • Event streaming
    • Observability
    • Payments & ledgers
    • Leadership

  7. iflix

    Principal Engineer

    Before that: Staff Engineer (2019), Engineering Lead (2016–19)

    Sep 2016 – Mar 2021

    Melbourne / Remote, Australia

    part of Tencent WeTV since Jul 2020

    Video streaming across Southeast Asia, acquired by Tencent WeTV.

    • Billing migration. Led the move from a third-party payment aggregator to in-house billing for 10m subscribers, integrating directly with a payment gateway and 8 telco billing partners, on a live platform.
    • Identity and payments migration to WeTV. Owned the integration of accounts, auth, subscriptions and payments with WeTV after the Tencent acquisition. Moved 60m accounts and 10m subscriptions with no stop in service (Go, SNS, SQS).
    • API consolidation. Set the direction for client-facing APIs. Got agreement across teams, then deprecated the legacy APIs and moved clients to one GraphQL platform (Apollo, gRPC, Redis, Akamai).
    • Identity service. Decommissioned most of the legacy account APIs and merged the rest into one service. Drove adoption of a new API version with better security and performance.
    • Data streaming. Led the design and delivery of the user analytics events pipeline: 1.2 billion events per day (Kinesis, Python, S3, Parquet). Built the first content recommendation system on it (Scala, Spark ML, EMR).
    • Production engineering. Drove CI/CD, Kubernetes, observability (Datadog, then Prometheus/Grafana), API CDN, gRPC and feature-flag work. Incident commander for major incidents; ran post-incident reviews.
    • Hiring. Ran technical interviews and set the hiring bar for engineers.

    60maccounts moved to WeTV, no downtime

    • Go
    • Python
    • Scala
    • AWS
    • Kubernetes
    • CI/CD
    • APIs
    • Event streaming
    • User analytics
    • Observability
    • Identity
    • Payments & ledgers
    • Leadership

  8. Envato

    Senior Developer

    May 2013 – Sep 2016

    Melbourne, Australia

    Creative marketplaces and online learning.

    • Built the tutsplus.com rewrite, including subscriptions with Recurly, Braintree and PayPal.
    • Led the front-end stack of Envato Elements (React, Redux). Drove AWS autoscaling, Packer, Ansible and CI/CD work.

    tutsplus.comrewrite shipped

    • Ruby
    • React
    • AWS
    • Infrastructure as code
    • CI/CD
    • Payments & ledgers

  9. PlayUp

    Senior Developer

    2011 – 2013

    Melbourne

    Web services and content apps, web and mobile. Rails, Node.js, PostgreSQL, MongoDB, AWS.

  10. List Globally

    Senior Developer

    2010 – 2011

    Melbourne

    Real estate listings aggregation and distribution. Rails, MySQL.

  11. LewatMana.com and ONEHUB Technology

    Co-founder and CTO; technical lead

    2008 – 2009

    Jakarta

    Road traffic monitoring; quantity surveying for home builders. Python, Django.

  12. Rice Studios

    Software Developer

    2006 – 2007

    Melbourne

    Email marketing app and its Linux servers. PHP, PostgreSQL, Postfix.

  13. Master of Science in Internetworking, 2006

    University of Technology Sydney

    Sydney

    High Distinction

  14. Seatech

    Consultant

    2001 – 2004

    Jakarta

    Sysadmin, networks and early web hosting.

  15. Bachelor of Computer Science (Honours), 2001

    University of Manitoba

    Winnipeg

AI / LLM

Practice

How I build with LLMs

  • Every model call through OpenRouter, with the model chosen per request
  • An OpenTelemetry span per model call, with GenAI semantic conventions
  • Every call kept in a ledger with billed cost and tokens, failures included
  • Hybrid retrieval, and async generation through Pub/Sub
  • Hand-checked answer sets with automatic scoring to drive prompt and model changes
  • MCP servers with OAuth 2.0 and per-user consent

Daily driver

Agent workflow

Builds with coding agents every day: Amp and Claude Code on Claude and GPT models, other models through OpenRouter, and local models in LM Studio. Agents run in isolated containers on Amp orbs and runners, with a maintained set of custom agent skills.

Toolbox

Tools and models

Tools

  • Amp
  • Claude Code
  • OpenRouter
  • LM Studio
  • Custom agent skills

Models

  • Claude
  • GPT
  • Gemini
  • Jev
  • Local models in LM Studio

Skills

Languages

  • Go Go services, shared libraries, MCP servers
  • TypeScript TypeScript, Node.js, Next.js
  • React React, Redux, Next.js
  • Ruby Ruby on Rails
  • Python Python, Django
  • Scala Scala, Spark ML

Platform

  • AWS EKS, SQS, SNS, EventBridge, Kinesis, Firehose, Aurora, EMR
  • GCP GKE, Cloud SQL, Pub/Sub
  • Kubernetes Kubernetes, Helm, ArgoCD, Linkerd
  • Infrastructure as code AWS CDK, Terraform/OpenTofu, Packer, Ansible
  • CI/CD GitHub Actions, Buildkite, ArgoCD, paved roads, service templates
  • Observability OpenTelemetry, Prometheus, Grafana, Datadog, Honeycomb, New Relic

Systems

  • APIs ConnectRPC, gRPC, GraphQL (Apollo)
  • Event streaming Kinesis, EventBridge, SNS/SQS, Firehose, Pub/Sub
  • Identity OAuth 2.0/OIDC, Ory Kratos
  • Payments & ledgers Stripe, Recurly, Braintree, PayPal, telco billing

Data

  • PostgreSQL PostgreSQL, Aurora
  • DynamoDB DynamoDB
  • TigerBeetle Double-entry ledgers on TigerBeetle
  • ClickHouse ClickHouse, Parquet on S3, Athena
  • User analytics Amplitude, Statsig

Practice

  • LLM systems Claude, OpenAI, Gemini, OpenRouter, Jev, local models in LM Studio, retrieval, evaluation, MCP
  • Leadership Architecture direction, mentoring, hiring bars, incident command

About

Ronny Haryanto, smiling, in glasses and a denim shirt

I’ve been a sysadmin, a consultant, a software developer, a data engineer, a site reliability engineer, a platform engineer, an engineering manager, and a staff and principal engineer.

I’ve worked on products in web hosting, email marketing, insurance, road traffic monitoring, construction, real estate, creative marketplaces, online learning, video streaming, workforce planning, healthcare and consumer banking.

I’m based in Melbourne. When I’m not working or spending time with my family, I enjoy cooking, coffee, music, and learning languages. I’m currently learning Japanese.

Strengths

  • AI / LLM
  • Paved roads for engineers
  • Live migrations
  • Event and streaming platforms
  • Payments and ledgers

Languages

English, Bahasa Indonesia, 日本語 (basic)