At the very beginning of the text , My current resume is available here.
I am Rayer Tung, an AI-native Staff Backend and Platform Engineer working primarily with Go, Java, and Rust. I focus on turning closed legacy systems and manual workflows into programmable, scalable, and reliable production platforms. I keep a hands-on scope across architecture, implementation, migration and rollout, production incident work, and multi-team adoption.
Career highlights
- AI-native engineering : Led a department-selected proof of concept on enterprise AI knowledge systems and incident workflow exploration. The work stayed source-grounded and focused on inspection and trust, with incident triage and remediation remaining human-reviewed.
- Legacy CORBA compatibility platform: At TSMC I converted nearly 1,000 CORBA IDL operations into HTTP endpoints, OpenAPI specs, and Go/Java SDKs. This closed integration gap for consumers while maintaining the legacy compatibility boundary, and it became a production integration path adopted by three internal teams.
- Automated migration platform for production VMs: Built and shipped an automated rolling migration for 1,400+ production VMs. A typical 6–7-node migration moved from about two engineer-days to roughly 1–2 largely unattended hours, with more than 99% first-run success and automatic retry handling for transient failures.
- Experiment integration modernization: At Binance I owned the RFC, API contract, server integration layer, and Go/Java/Rust SDKs that turned a web-only A/B testing integration path into a programmable platform. The platform handled about 40K QPS and was adopted by 7+ business units.
- High-scale recommendation and serving: At Appier I owned a recommendation and advertising serving platform at 20K–60K QPS and re-architected core serving behavior into Go on GKE. Matched-period billing comparisons showed an estimated 65% reduction in product-attributed cloud costs.
LLM Wiki Cloud
I am building LLM Wiki Cloud end to end as an AI knowledge workspace. It turns Markdown and web sources into structured, searchable, source-linked knowledge and coherent topic-level entries.
In enterprise knowledge workflows, fragment-level RAG often retrieves passages without a coherent system overview. I built concept-oriented compilation across documents so teams get organized topics with source traceability.
I own this end-to-end: product scope, architecture decisions, and implementation.
- Multi-tenancy and tenant isolation
- BYOK
- Markdown and web ingest
- Ingest, compile, and publish workflow
- Structured knowledge generation and search
- Retrieval with source traceability
- Publicly browsable demo content
Try the public demo (choose Try Demo on the sign-in screen).
The aim is not merely answer generation. The goal is knowledge outputs people can understand, verify, and trust.