- Designed a multimodal VLM system integrating speech, facial expression, and text for emotion-aware interaction with elderly users
- Built a fusion pipeline using Whisper, wav2vec, and ResNet-18 to address cross-modal inconsistency in emotional signals
- Iteratively refined prompting strategies to guide the LLM toward emotionally appropriate responses under ambiguous inputs
- Developed RGB-based and X-ray-based defect detection models for industrial pipe welding inspection
- Trained ConvNeXt V2 on balanced RGB datasets and YOLOv8 on highly imbalanced X-ray datasets to learn robust visual representations
- Integrated defect detection as an upstream module for data-driven rework decision-making and scheduling automation
- https://github.com/Automated-Pipe-Welding-Rework-System
- Designed an LLM-based reasoning agent that evaluates company-authored OKRs using structured scoring (1–5), justification, and corrective suggestions
- Applied chain-of-thought prompting and self-consistency to improve reliability of reasoning outputs
- Integrated retrieval-augmented generation (RAG) with company-specific data and explored parameter-efficient fine-tuning (LoRA, quantization) on LLaMA-based models
- https://github.com/hayeon7898/OKR-AI-Agent
- Conducted mentored research under guidance from IonQ
on adaptive hybrid quantum–classical machine learning with optimized quantum data embedding
on IonQ trapped-ion hardware - Designed hybrid optimization pipelines under limited measurement (probing) budgets, focusing on quantum state estimation and state transformation
- Validated quantum circuits through iterative execution on simulators and real quantum hardware, emphasizing gate-level optimization
- https://github.com/IQ-rhythm/IonQ-Encode
- Challenger Level Award
- hosted by Quantum Information Research Support Center
- 1st Place
- hosted by MIT(Massachusetts Institute of Technology)
Frontend Database
- Implemented the core organization tab using React Native
- Designed PostgreSQL schemas and optimized queries with Supabase, including index-level performance tuning
- https://github.com/Veloci-Lab/dearday
Backend Database Infra
- Built backend foundations and service skeletons, including initial JDA setup
- Implemented approximately half of the core APIs supporting appointment workflows
- Integrated Kakao Skill and Event APIs to implement chatbot-based appointment flows and automated notifications
- https://github.com/hayeon7898/Scheduly
Frontend Backend Database
- Full-stack developer building a travel itinerary planning service
- Built interactive itinerary planning UI with Kakao Map, day-by-day scheduling, drag-and-drop place management, and protected routes
- Implemented itinerary management, place search, recommendations, and JWT authentication using FastAPI and PostgreSQL
- https://github.com/hayeon7898/BeTrip-Frontend, https://github.com/hyun7931/BeTrip-Backend
Frontend Backend Database Infra
- Full-stack developer in a two-person team building a QR-based umbrella rental and return service
- Implemented rental logic, reward systems, and real-time state management using React Native, Nest.js (TypeScript), and MySQL
- https://github.com/hayeon7898/bbosong-front, https://github.com/hayeon7898/bbosong-back
- A gyro-based mobile game featuring satisfying shotgun combat and risk-taking gameplay against enemies.
- Developed with Unity and tested on mobile devices.
- https://github.com/hayeon7898/Gyro-Shotgun-Game
- Participated in a club-wide competition to recreate the Crazy Arcade game
- Developed the game for PC using Unity and C#
- https://github.com/hayeon7898/Copy-Crazy-Arcade
- Developed an NPU accelerator for Edge LLM inference based on Rocket Chip.
- Implemented a RoCC interface for custom instruction handling and autonomous execution using Task Descriptors.
- Designed and implemented the Control Unit, developed the hardware architecture using Chisel (Scala).
- https://github.com/hayeon7898/lzero-rtl
- Developed various embedded applications on Cortex-M0+, including RC car control and register-level peripheral control.
- Implemented GPIO, timers, PWM, interrupts, and communication interfaces.
- Worked with Thumb ISA and low-level hardware control.
- Developed the Teleoperation module for an exoskeleton-based robotic system.
- Processed and filtered joint angle data received from the exoskeleton.
- Published target joint angles to the robot control module using ROS 2.
- https://github.com/hayeon7898/avatar_2026
- Implemented a CPU scheduler at the Linux system call level, adding a custom system call for scheduling policy selection and execution.
- Implemented and compared FCFS, SRTF, Round Robin, and Priority with Preemption scheduling policies.
- Developed kernel-level process management using queues, process control structures, and spinlocks.
- Tested scheduling behavior and response times in an Ubuntu environment running on VirtualBox.
- https://github.com/hayeon7898/linux-kernel-cpu-scheduler
AI / Machine Learning
Vision & Multimodal
Speech
Frontend / Backend
Database
Infra
Tools
- Email: hayeon22@korea.ac.kr