Skip to content
View Jang1J's full-sized avatar

Block or report Jang1J

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Jang1J/README.md

Jang1J

AI Systems Builder

신뢰할 수 있는 AI 시스템, 멀티에이전트 파이프라인, 재현 가능한 ML 실험을 만듭니다.

Python PyTorch LightGBM gRPC Kafka

Featured: Flephant · All repositories

About

모델의 정확도만큼 운영 안전성, 데이터 시점 정합성, 실패 시 복구 가능성을 중요하게 생각합니다. 연구 아이디어를 코드로 옮기는 데서 멈추지 않고, 계약·검증·관측 가능성을 갖춘 시스템으로 완성하는 작업에 관심이 있습니다.

  • Multi-agent AI orchestration and decision systems
  • Financial ML, ranking, reinforcement learning, and risk controls
  • Reproducible evaluation, contract testing, and paper-safe operations
  • Python 중심의 AI와 Java/TypeScript 기반 서비스 경계 통합

Featured Project

KOSPI 30종목을 1분 주기로 분석하는 멀티에이전트 Decision OS입니다. 장중의 저지연 정량 경로와 이벤트 기반 LLM 경로, 장마감 후 모델 검증 경로를 하나의 paper-safe 시스템으로 연결했습니다.

Area Implementation
Decision pipeline LightGBM ranking → PPO allocation → portfolio manager → FDA approve/veto
Agent system Quant, News, Risk, Debate, FDA, Backtest agents
Reliability 18 API contracts, PIT-safety, replay gates, audit trail
Integration gRPC, Kafka, REST, STOMP
Result Capstone project A+, 2,221 tests collected

프로젝트 자세히 보기 →

Toolbox

Layer Stack
AI / ML Python, PyTorch, LightGBM, scikit-learn, reinforcement learning
Agent Systems Multi-agent orchestration, RAG, LLM routing, typed contracts
Services gRPC, Kafka, REST, STOMP, Spring Boot
Frontend TypeScript, React, Vite
Reliability pytest, replay validation, PIT-safety, audit logging, circuit breaker

More Work

Project Focus
GPT-2 Sonnet Generation 구조 제약 디코딩과 재현 가능한 생성 품질 평가
Food Regulation Chatbot 식품 법규 문서를 위한 RAG 기반 질의응답
Food Safety Mailer 글로벌 식품 안전정보 자동 수집 및 이메일 전달
Nuclear AI NAS 원전 사고 진단을 위한 AI 모델 탐색

GitHub Activity

Jang1J GitHub profile summary Jang1J contribution activity graph

Engineering Principles

correctness before cleverness
evidence before confidence
fail closed on unsafe transitions
make experiments reproducible

Pinned Loading

  1. 26_nuclear_ai_NAS 26_nuclear_ai_NAS Public

    AI-based nuclear accident diagnosis with reproducible model search experiments

    Python

  2. Flephant Flephant Public

    KOSPI 30-stock, 1-minute multi-agent Decision OS with LightGBM, PPO, explainable gates, and paper-safe validation.

    Python

  3. food-regulation-chatbot food-regulation-chatbot Public

    RAG-based Q&A for food regulations with source-grounded answers

    Python

  4. food-safety-mailer food-safety-mailer Public

    Automated global food-safety monitoring and email delivery using public data APIs

    Python

  5. gpt2-sonnet-generation gpt2-sonnet-generation Public

    GPT-2 sonnet generation with structured decoding, evaluation, and reproducible analysis

    Jupyter Notebook

  6. KANANA-429 KANANA-429 Public

    Reproducible comparison of Kanana-2 2.1B Dense and 30B MoE models

    Jupyter Notebook