Stat·1h ago
35%
post-search bounce rate fell from
AI Fluency
Q&A
No questions yet
Nothing posted yet.
Projects
C
Career Mate 커리어 멘토 에이전트
Multi-agent career mentoring service at Wanted
Six workflows and a multi-agent system covering mentoring, mentor recommendation, resume diagnosis, and nudge message generation. Resume diagnosis raised data quality for 11,626 cumulative users.
번
번역 모델 경량화
Compressing MBART for service deployment
Built a half-depth MBART with knowledge distillation, then applied PTQ and attention-based pruning. Exported to ONNX as a static graph for serving.
Q
Question Difficulty Estimation (Applied Sciences)
Attention-based question difficulty model for QA
Co-first-authored a paper on estimating question difficulty with an attention model for question answering.
Experience
(
AI 기술팀 정규직
(주)원티드랩Nov 2025 — Present · Seoul, South Korea
Replaced the BM25 position search engine with a SPLADE + Dense + Lexical 3-leg hybrid search, tuned via A/B tests. Post-search bounce rate fell from 35% to 28% and search-driven application conversion rose from 6.7% to 8.4%. Also built the career mentor multi-agent service, where resume diagnosis raised user resume levels, and migrated legacy position agent code to LangChain and LangGraph.
Post·1h ago
As an AI Engineer at Wanted Lab in Seoul, I've focused on improving our core search capabilities and developing new agent-based services. I replaced our BM25 position search engine with a hybrid system combining SPLADE, dense, and lexical search. Through A/B testing, this change reduced the post-search bounce rate from 35% to 28% and increased search-driven application conversion from 6.7% to 8.4%.
I also built the career mentor multi-agent service. This service uses resume diagnosis to help users improve their resume levels. Additionally, I migrated our legacy position agent code to LangChain and LangGraph, streamlining our agent infrastructure.
웨
AI Engineer → AI Engineer Leader
웨이커코리아 주식회사Oct 2022 — Aug 2025 · Seoul, South Korea
Built a multi-agent AI chat API that uses MCP tools grouped per agent, LLM orchestration, and a GuardRail module blocking sensitive identifiers. Developed a financial stock analysis agent using Plan-and-Act workflows, a Llama-Index QA agent that validates extracted SEC filing data, and a Kafka-based news pipeline with translation and summarization APIs. Earlier work covered SEC filing extraction with LLM and multimodal models, plus model compression for translation (distillation, pruning, quantization, ONNX).
Article·1h ago
Article: AI Engineer → AI Engineer Leader at 웨이커코리아 주식회사
Post·1h ago
At Waker Korea, I focused on developing AI agents and LLM-driven systems. One key project was a multi-agent AI chat API. It used LLM orchestration, grouped MCP tools for each agent, and a GuardRail module to block sensitive identifiers. I also built a financial stock analysis agent with Plan-and-Act workflows and a Llama-Index QA agent for validating extracted SEC filing data.
My work also included a Kafka-based news pipeline, integrating translation and summarization APIs. Earlier, I worked on SEC filing extraction using LLM and multimodal models. I also explored model compression techniques like distillation, pruning, quantization, and ONNX for translation models.
(
NLP Developer 정규직
(주)디앤아이파비스Jul 2021 — Oct 2022 · Seoul, South Korea
Built a BM25 patent document search engine on self-hosted Elasticsearch. Domain-tuned Sentence-BERT for problem and solution sentence matching, improving accuracy by 1.2% over vanilla BERT. Developed an LDA-based keyword extraction and clustering API for patent documents.
Stat·1h ago
1.2%
accuracy improvement
Post·1h ago
During my time as an NLP Developer at D&I Fabis in Seoul, I focused on building practical applications for patent documents. One key project was developing a BM25 patent document search engine, which I set up on a self-hosted Elasticsearch instance.
I also worked on improving semantic search. I domain-tuned Sentence-BERT for matching problem and solution sentences, which boosted accuracy by 1.2% compared to using vanilla BERT. Additionally, I created an API for keyword extraction and clustering in patent documents, using an LDA-based approach.
네
인턴 연구원
네이버 주식회사Feb 2021 — May 2021 · Seoul, South Korea
Researched training methods for ColBERT in information retrieval. Used Learning-to-Augment to generate hard negative samples and improved the Triplet Loss training approach.
Article·1h ago
Article: 인턴 연구원 at 네이버 주식회사
Post·1h ago
During my internship at Naver in 2021, I focused on improving training methods for ColBERT in information retrieval. This involved exploring how to generate hard negative samples using Learning-to-Augment.
My work aimed to refine the Triplet Loss training approach. It was a valuable experience to contribute to research in this area within a company like Naver, known for its search and AI advancements in Seoul.
Credentials
No credentials yet
Nothing posted yet.
Recommendations
No recommendations yet
Nothing posted yet.
Skills
LangChainLangGraphLlama-IndexAgentic AIMulti-AgentMCPPrompt EngineeringGuardRailHybrid SearchSPLADEBM25OpenSearchElasticsearchFAISSRerankingA/B Testing