Skip to main content

Projects & Dev Logs

Select a project to explore dev logs and technical milestones.

Active · 23 PostsCapstone Design / AI Chatbot

Maeumjigi

AI Suicide Prevention Counseling Chatbot Dev Log

Real-time crisis & depression detection with compassionate counseling. Llama 3.1 fine-tuning (2:1 emotional-to-daily ratio), Safety Wrapper guardrails, and AWS EC2 GPU deployment.

#PyTorch#Llama 3.1#Safety Wrapper#FastAPI#AWS EC2#CapDe
Explore Dev Logs (23 Posts) →
Active · 5 PostsComputer Vision / ITS

Blackbox ANPR & Lane Intrusion

Real-Time Lane Intrusion Detection & Super-Resolution ANPR System

Real-time lane intrusion detection with IoU, blurry license plate reconstruction via EDSR_x4 super-resolution, and temporal majority voting EasyOCR pipeline across 11 iterations (v01 to v11).

#OpenCV#YOLOv5#EDSR (SR)#CSRT Tracker#EasyOCR#PyTorch
Explore Dev Logs (5 Posts) →
Active · 5 PostsReinforcement Learning / Robotics

RL Q-Learning

Webots Q-Learning Autonomous Robot Project

Q-Table optimization and autonomous navigation in Webots simulation environments. Continuous state discretization (648 bins), socket IPC communication, and Sim-to-Real hardware transfer.

#Python#Webots#Q-Learning#NumPy#Socket IPC#Arduino
Explore Dev Logs (5 Posts) →
Active · 1 PostWeb Security / DevSecOps

Security & Web Engineering

Static Site Security Hardening & Supply Chain Defense

Threat modeling 45 npm audit vulnerabilities in static sites, git secret scanning, PII protection, Referrer Policy headers, and automated security verification pipelines.

#npm audit#SSG Hardening#DevSecOps#Threat Modeling#CI/CD
Explore Dev Log (1 Post) →
PlannedDeveloper Tools / AI Agents

AI Research & Tools

Agentic AI & Custom Dev Workflows

Customizing Antigravity CLI, agent orchestration, prompt engineering, and building developer productivity tools.

#Antigravity#Agentic AI#Prompt Engineering#TypeScript

All Engineering Logs & Articles Index

From 45 npm Vulnerabilities to Zero-Risk: Why You Shouldn't Blindly Run npm audit fix --force on Static Sites[Vision AI] Korean ANPR OCR Engine, Regex Normalization & Temporal Majority Voting (v01 to v11 Evolution) (#5)[Vision AI] Overcoming Blurry Blackbox Artifacts: Deep Super-Resolution (EDSR vs FSRCNN) & Multi-Frame Image Stacking (#4)[Vision AI] Hierarchical Object Detection: YOLOv3 Vehicle ROI, YOLOv5 Plate Detector & CSRT Tracking (#3)[Vision AI] Dynamic Lane Boundary Extraction & Geometric IoU Intrusion Criterion (#2)[Vision AI] Blackbox Lane Intrusion & Super-Resolution ANPR: 6-Stage End-to-End Pipeline & Real-Time Constraints (#1)[Maeumjigi Capstone Dev Log #23] Safety Wrapper & 1st Semester Final Presentation[Maeumjigi Capstone Dev Log #22] 4th Fine-Tuning Complete! Architecture Design & Final Report[Maeumjigi Capstone Dev Log #21] Optimizing Data Mix Ratios and Starting the 4th Fine-Tuning[Webots Robotics] Sim-to-Real: Hardware Transfer via Arduino Uno, HC-06 Bluetooth & ArUco Vision Tracking (#5)[Maeumjigi Capstone Dev Log #20] 3rd Fine-Tuning Complete & Backend Feature Enhancements[Webots Robotics] Deterministic Evaluation via evaluate_qlearning & Benchmark vs. DQN and SARSA (#4)[Webots Robotics] Conquering Local Minima in 12-Wall Obstacles: Multi-Seed Headless Simulation (#3)[Webots Robotics] Tabular Q-Learning Implementation & Reward Shaping Tuning (#2)[Maeumjigi Capstone Dev Log #19] Automating Reports and Backend Database (DB) Troubleshooting[Webots Robotics] Q-Learning Autonomous Navigation: 648-Bin State Discretization & Socket Architecture (#1)[Maeumjigi Capstone Dev Log #18] Expanding Inference Tokens and Comparing 2nd Fine-Tuning Results[Maeumjigi Capstone Dev Log #17] 2nd Fine-Tuning Complete & The Trap of the Automation Pipeline[Maeumjigi Capstone Dev Log #16] Fatal Mistake, 2nd Fine-Tuning, and React Native App Launch[Maeumjigi Capstone Dev Log #15] Visualizing Fine-Tuning Results of 5 Models & Local Environment Setup[Maeumjigi Capstone Dev Log #14] All Fine-Tuning Completed and Parallel Inference (Eval) Pipeline Built[Maeumjigi Capstone Dev Log #13] Limits of 70B Training and the Pivot to Mid-Size Models (HCX, Qwen)[Maeumjigi Capstone Dev Log #12] 70B Massive Model Fine-Tuning and GPU Memory (OOM) Struggle[Maeumjigi Capstone Dev Log #11] Evaluating M6 Model Trained on Synthetic Data (v5)[Maeumjigi Capstone Dev Log #10] Lifting Application Programming Interface (API) Limits and Resuming Data Synthesis[Maeumjigi Capstone Dev Log #9] Llama 3.1 Fine-Tuning Error Struggle and Week 6 Mid-term Presentation[Maeumjigi Capstone Dev Log #8] Application Programming Interface (API) Bill Bomb Incident and Preparing Colab Fine-Tuning[Maeumjigi Capstone Dev Log #7] Synthetic Data (v5) Generation Optimization and Frontend Integration[Maeumjigi Capstone Dev Log #6] Preprocessing Final Report Automation and Frontend Kickoff[Maeumjigi Capstone Dev Log #5] Purifying 100,000 Emotion Data Points and the Pressure of Application Programming Interface (API) Costs[Maeumjigi Capstone Dev Log #4] Full-Scale Fine-Tuning Data Preprocessing[Maeumjigi Capstone Dev Log #3] Alarm App Build Test and Fine-Tuning Preprocessing Kickoff[Maeumjigi Capstone Dev Log #2] Initial Alarm App Planning and Testing[Maeumjigi Capstone Dev Log #1] The First Steps and Planning