h
hykura

Tiger Vo

@hykura

Fullstack Web Development

Vietnam
Vietnamese, English
About me
I am a Software Engineer specializing in backend development with Golang, microservices, and cloud infrastructure. I have experience building scalable systems using gRPC, Kafka, and Kubernetes, and developing AI-driven applications with RAG and LLMs. I am passionate about high-performance architecture and automated CI/CD workflows.... Read more

Skills

h
hykura
Tiger Vo
Offline • 
Average response time: 1 hour

See my services

API & Integrations
I will develop rest apis using golang and postgresql
Full Stack Web Applications
I will build full stack web applications using react and nodejs

Portfolio

Work experience

VNG

Software Engineer

VNG • Full-time

Jul 2026 - Present • 4 mos

• Built an end-to-end encrypted 1-1 chat platform of 7 microservices using Golang, gRPC, Cassandra, Kafka, and Redis, designed so the server only relays ciphertext and never accesses plaintext messages. • Implemented Signal-style E2EE with X3DH key agreement over published prekey bundles (identity, signed, and one-time prekeys) verified by Ed25519 signatures, relaying Double Ratchet payloads as an opaque header and ciphertext to achieve per-message forward secrecy. • Delivered messages in real time with a router service (Consumer) that consumes message events from Kafka, looks up which pod holds the receiver’s live connection in a Redis registry, and calls that pod over gRPC to push the message down its WebSocket. • Implemented per-message status tracking for sent, delivered, and read, driven by ACKs from the receiver client. • Wrote message rows and their outbox event in one Cassandra logged batch (Transactional Outbox), then relayed events to Kafka at-least-once with deterministic event IDs for consumer-side deduplication. • Sharded the outbox across 256 bucket partitions claimed by workers in parallel, with autoscaling driven by backlog size and Snowflake IDs as a time-ordered resume cursor. • Made a retried send resolve to the same message instead of a duplicate by claiming each client idempotency key in Redis with an atomic SET NX, with deadlines carried down the call chain to cut off late attempts. • Deployed to a 3-node Kubernetes cluster via GitOps with Argo CD and GitLab CI for high availability and hori zontal scaling, with Prometheus, Grafana, and Loki for metrics and centralized logs. • Built a load-testing tool simulating concurrent chat clients over WebSocket; the system sustained ~3,000 realtime msg/s at p99 under 500ms with zero message loss