Bongkeun Shin

- Currently in research to improve the performance of Self-Reflective-RAG through finetuning LLM models. - Conducted topic modeling with BERTopic for documents from search engine to enhance the accuracy of the search engine. - Finetuned classification model based on Sentence Transformer for search portal. - Developed logging program provide training data and re-indexing CLI to reindex Elasticsearch indices. - Finetuned model for 3D reconstruction based on NeRF, NSVF and instant NeRF for better performance.

- As a Developer, designed and developed the bug bounty web platform for white hackers to report vulnerabilities and corporate users to view reports, focusing on protecting platform from white hackers. Developed in-browser virtual desktop service on portal which could be accessed with Guacamole and Kali Linux. - As a Co-founder, promoted the service to potential customers by conducting sales activities. Gathered requirements and feedback from white hackers, security personnel and C-level executives.

- As a Consultant, established and configured business process with the process innovation personnel from Samsung C&T and Samsung Heavy Industry, to provide value to the system users. As a ABAP Developer, developed ERP system with ALV Reports, IDOCs, OSS Notes, Smartforms, EDIs, User-Exits, SQL Enhancements.

Final Project: Comparing both cost-free and paid models for Chatbot in experiments of finetuning pretrained Llama-2-13b, flan-t5 and GPT-3.5, and conducting Retrieval Augmented Generation on finetuned Llama-2-13b and GPT-4 models. Verified Llama-2 can perform comparable to GPT-4 on RAG experiments