Soohan Park 박수한

loading...

Designed and built the ECVRF-based randomness service used for item enhancement in MapleStory N, and own its deployment and ongoing operations. Implemented the flow from requesting randomness to serving it to game servers and publishing it on-chain after expiry. In 2024, added in-memory caching of the current seed, reducing GetSeed P95 latency from 66.35ms to 1.52ms in a DEV comparison test at 350 RPS, then deployed the improvement through LIVE. In 2025, implemented and tested seed and proof encryption while preserving compatibility with existing plaintext records.

Built Merkle Processor, which turns NXPC Fusion basket and reward data into Merkle trees for on-chain registration, and implemented reward registration for NXPC airdrops. For the Teller (Warp) backend, implemented fee calculation, request expiry settings, and bridge integration for cross-chain asset transfers. Connected gRPC APIs, database schemas, and contract calls, handling resubmission of batched reward data and retries after on-chain failures.

Separated signers by action and asset type and redesigned the transaction submission service, with load testing in STAGE and deployment to DEV, TEST, QA, and STAGE. Built a Kafka message replay tool and fixed duplicate token payouts after pod restarts by checking each transaction's status before resending. In 2026, designed and implemented a high-availability backend using leader election and database-managed state, and carried out local HA verification and performance testing.

Worked with the infrastructure team to prepare the production EKS environment, from provisioning requests through deployment manifests, image registries, deployment jobs, database connections, and service startup. Adjusted preStop hook use to each service's pod configuration, implemented health checks, ran failover tests, and improved gRPC keepalive and DNS resolution. Introduced Datadog to the team and expanded it into cross-service tracing, alerts, and dashboards to follow latency and errors through each processing stage alongside resource usage. In 2026, used a year of Datadog usage data to adjust pod counts, CPU, and memory allocations, keeping headroom for critical transaction services and checking service health with colleagues after the LIVE rollout.

Built three kinds of Avalanche and subnet archive nodes for historical block and state queries and accounting audit data collection. Handled EC2 configuration, node registration, sync verification, and containerization, and set up deployment environments for the archive crawler that stores block data in S3. Investigated sync delays and storage issues, proposing storage changes based on cost and reliability. For the Fortuna, Granite, and Helicon upgrades, coordinated schedules with the relevant teams, updated archive nodes, and verified the mainnet rollout.

Worked with performance testing and infrastructure teams to test the Avalanche subnet and blockchain backend, owning test plans, environment setup, execution, and reports for each round. Tested hundreds of TPS of transaction load and tens of thousands of RPS of read load, both separately and together, with runs spanning several days to check latency and service health. Traced unprocessed transactions to investigate backend pipeline bottlenecks and RPC overload, feeding the findings into configuration tuning and operational baselines such as production RPC node counts. Also built a gRPC load generator so the team could validate changes directly.

Built Nexus, an internal toolkit for decoding transaction input data, looking up contract and domain information, and running deployment jobs. Its Rundeck Dashboard addresses repetitive work and missed objects when running deployment jobs individually across environments. The interface lets developers select objects and versions in one place and run the jobs together; I deployed it to TEST, QA, and STAGE. An earlier GitLab CI automation attempt had not gained traction because it was inconvenient to access, so I rebuilt the workflow as a web tool the team could use directly.

Developed ERC-721 and ERC-1155 NFT smart contracts with proxy-based upgrades, and automated deployment and management with Hardhat and Ethers.js. Built tools for holder snapshots, ownership verification, IPFS uploads, and a randomness oracle. Alongside fully on-chain NFTs, researched splitting and assembling SVG data across multiple contracts and co-authored a paper on the tradeoff between contract size limits and deployment cost.

Built ChainOne, a private network on AWS Managed Blockchain and Hyperledger Fabric, and integrated it with CJ Logistics' pharmaceutical supply chain system. Developed Go chaincode and Java/Spring APIs to record and query receiving, shipping, and transport events by pallet identifier (SSCC). The system separates current state from the full change history for tracking and verification; I owned the network configuration, chaincode, and API development and co-authored the related paper.

Contributed to Hana Life's mobile web and app renewal, supporting frontend development for insurance diagnosis, recommendations, and LifeCheck. Wrote the PIST app's information architecture and screen specifications, and helped manage schedules and deliverables. Later worked on Open APIs that bring Hana Life's insurance information and recommendations into Hana Bank's app, implementing Java/Spring APIs and database queries from existing designs and running integration tests across affiliates.