Application Position
GoLang Engineer
First Name
Derrick
Last Name
Luo
Telegram
@derrick_rose121
How did you find out about this job?
jobsdb
Link to Portfolio
www.linkedin.com/in/derrick-luo-69a26931a
Link to LinkedIn Profile
www.linkedin.com/in/derrick-luo-69a26931a
Short Description
Dear Holdex Team,
I am applying for the GoLang Engineer role. Three years of my five-year backend career have been Go-focused, building distributed systems that handle financial data in production — and the ownership model you describe, where engineers define and drive their own work without managers, is how I already operate.
Go in production, at the systems-design level. At Hongkong Nova I develop an aggregated payment and digital wallet platform: 10+ go-zero microservices behind a unified gateway, gRPC between services, Etcd for service discovery, Kafka for async decoupling, serving acquiring, wallets, clearing, settlement and reconciliation across 10+ payment channels. Financial data is unforgiving, so most of the interesting work is in guarantees rather than features:
Consistency. A transaction must persist and its downstream message must be delivered, without loss and without double-counting — and a distributed transaction on the hot path is too expensive. I implemented eventual consistency via a local message table (Outbox) with idempotency keys: the write and message enqueue commit atomically in one local transaction, consumers deduplicate on the key. Result in production: accounting error-free, zero capital-loss incidents, reconciliation discrepancy inside 0.01%.
Load behaviour. Splitting payment callbacks into synchronous and Kafka-buffered asynchronous paths means channel surges shed load rather than propagate; the core path holds at peaks of ~3,000 QPS without jitter.
Choosing boundaries. A channel adapter layer hides protocol differences behind one interface, so onboarding a new channel takes a day instead of three with zero core changes. When analytical queries started degrading the transaction database, I moved reporting to a StarRocks read/write-split instead of tuning queries indefinitely — 20s+ down to 1–2s.
Alongside this I built an enterprise AI Agent platform in Go (Hertz, Eino, Milvus) with two systems live in production, and independently developed its MCP Server — authentication, transport, toolset and metrics — exposing business capabilities as standard protocol tools for third-party clients. Designing a protocol-level integration surface that others build against is, I suspect, close in spirit to the network you are building.
On working asynchronously. I have owned modules end to end rather than tickets — scoping the problem, designing the approach, shipping it, and being the person accountable when it breaks. Concretely: I independently owned the Kafka async ingestion pipeline (hash deduplication, idempotency, retries, sharding) and the MCP Server, both from design through production. I write things down by default — design decisions and their trade-offs, not just what was built — and I raise problems early rather than at the deadline. Deciding what to do without being told is the part of your description I find most attractive.
Where I stand on web3. I should be direct: my production experience is fintech rather than blockchain. I have studied the space independently — consensus mechanisms, wallet key management, transaction models — but I would be claiming too much to call it professional experience. What I bring instead is distributed systems judgement built where correctness is enforced by money: idempotency, eventual consistency, replay safety, reconciliation. My read is that those translate closely to on-chain work, and I learn fast — I led a core-module migration from C++ to Go, and taught myself the Agent and vector-retrieval stack from scratch to ship two production systems.
I am based in Hong Kong, hold a BEng in Electronic Science and Technology, and work comfortably in English and Mandarin.
Thank you for your consideration. I would be glad to walk through the Outbox design or the MCP Server in as much depth as you would like.
Best regards,
Derrick (Luo Zihao)
Phone / WhatsApp: +852 97002038
Email: [email protected]
Derrick-Golang-EN.pdf
Application Position
GoLang Engineer
First Name
Derrick
Last Name
Luo
Telegram
@derrick_rose121
How did you find out about this job?
jobsdb
Link to Portfolio
www.linkedin.com/in/derrick-luo-69a26931a
Link to LinkedIn Profile
www.linkedin.com/in/derrick-luo-69a26931a
Short Description
Dear Holdex Team,
I am applying for the GoLang Engineer role. Three years of my five-year backend career have been Go-focused, building distributed systems that handle financial data in production — and the ownership model you describe, where engineers define and drive their own work without managers, is how I already operate.
Go in production, at the systems-design level. At Hongkong Nova I develop an aggregated payment and digital wallet platform: 10+ go-zero microservices behind a unified gateway, gRPC between services, Etcd for service discovery, Kafka for async decoupling, serving acquiring, wallets, clearing, settlement and reconciliation across 10+ payment channels. Financial data is unforgiving, so most of the interesting work is in guarantees rather than features:
Consistency. A transaction must persist and its downstream message must be delivered, without loss and without double-counting — and a distributed transaction on the hot path is too expensive. I implemented eventual consistency via a local message table (Outbox) with idempotency keys: the write and message enqueue commit atomically in one local transaction, consumers deduplicate on the key. Result in production: accounting error-free, zero capital-loss incidents, reconciliation discrepancy inside 0.01%.
Load behaviour. Splitting payment callbacks into synchronous and Kafka-buffered asynchronous paths means channel surges shed load rather than propagate; the core path holds at peaks of ~3,000 QPS without jitter.
Choosing boundaries. A channel adapter layer hides protocol differences behind one interface, so onboarding a new channel takes a day instead of three with zero core changes. When analytical queries started degrading the transaction database, I moved reporting to a StarRocks read/write-split instead of tuning queries indefinitely — 20s+ down to 1–2s.
Alongside this I built an enterprise AI Agent platform in Go (Hertz, Eino, Milvus) with two systems live in production, and independently developed its MCP Server — authentication, transport, toolset and metrics — exposing business capabilities as standard protocol tools for third-party clients. Designing a protocol-level integration surface that others build against is, I suspect, close in spirit to the network you are building.
On working asynchronously. I have owned modules end to end rather than tickets — scoping the problem, designing the approach, shipping it, and being the person accountable when it breaks. Concretely: I independently owned the Kafka async ingestion pipeline (hash deduplication, idempotency, retries, sharding) and the MCP Server, both from design through production. I write things down by default — design decisions and their trade-offs, not just what was built — and I raise problems early rather than at the deadline. Deciding what to do without being told is the part of your description I find most attractive.
Where I stand on web3. I should be direct: my production experience is fintech rather than blockchain. I have studied the space independently — consensus mechanisms, wallet key management, transaction models — but I would be claiming too much to call it professional experience. What I bring instead is distributed systems judgement built where correctness is enforced by money: idempotency, eventual consistency, replay safety, reconciliation. My read is that those translate closely to on-chain work, and I learn fast — I led a core-module migration from C++ to Go, and taught myself the Agent and vector-retrieval stack from scratch to ship two production systems.
I am based in Hong Kong, hold a BEng in Electronic Science and Technology, and work comfortably in English and Mandarin.
Thank you for your consideration. I would be glad to walk through the Outbox design or the MCP Server in as much depth as you would like.
Best regards,
Derrick (Luo Zihao)
Phone / WhatsApp: +852 97002038
Email: [email protected]
Derrick-Golang-EN.pdf