Dahae Shin

Software Engineer student dedicated to improving skills through hands-on learning and development work. Demonstrates proficiency over multiple programming languages. Adept at using python through multiple school projects related to machine learning, data manipulation and visualization. Well-organized and collaborative team player with strong communication and analytical abilities.

• Design and prototype of knowledge-based agents
  Designed two intelligent agents (Agent A and B), Agent A to park a car at a parking lot and Agent B to control parking lots where the Agent A would navigate. The project covered both designing each agent's knowledge base (environment and rules) and programming to run a prototype.
  
• Developing web application for a smart city, ‘Bin Detective'
  Design and developed a web application to help the user easily find a nearby trash bin in a smart city environment. Data analysis was done using python using the open-source database with geo-locations then the visualization of the website was done by using JavaScript and html templates. For map visualization google API was used to pin the location.
  
• Food recipe recommendation program through image analysis
  Developed food recipe recommendation system using images of spices and herbs. Designed and developed a system enabling image-based input to food recipe recommender systems. Prototyped image classification system and recipe recommender systems by successfully training the system to result in relatively high-test accuracy with photos of new spices and herbs and recommended relevant recipes to what a user requested.
• Detection of stable gripping situation for slip prevention by a robot hand gripper
  Current graduation research is to detect non-deformable objects' stable gripping situation to prevent slipping by automatically handling gripping force in real-time. Gripping situations are demonstrated using a tactile sensor developed by the lab to collect data stable and unstable gripping situations. Utilizing the collected data, deep neural networks of 1DCNN, 2DCNN, LSTM and RNN have been implemented to achieve the best classification result. The best type of neural network will be applied to demonstrate the gripping situations using real objects (books, paper cups, plastic cups etc.) in real time. Aimed accuracy for this real time feedback system is 95%. The classification result will provide a feedback to control the robot gripper force will be implemented to complete the system.