Hi, I'm Kohki Hatori
BA/MS Computer Science Student
@ Boston University
Motivated computer science student with strong problem-solving and teamwork skills. Experienced in applying data-driven and software engineering approaches through internships and projects.
About Me
I am a Computer Science student at Boston University with a strong interest in full-stack development and applied machine learning. My academic and internship experiences have given me hands-on skills in building end-to-end systems, from designing databases and APIs to deploying production-ready cloud applications.
I enjoy exploring new technologies and turning ideas into practical solutions that create real impact. With a balance of technical depth, adaptability, and collaboration, I work effectively in both independent projects and cross-functional teams.
Projects
Backpacking Assistant Agent
Architected an AI-native travel platform using Next.js 15 and FastAPI, employing a LangGraph multi-agent system to orchestrate itinerary generation and real-time research. Integrates Perplexity AI and Gemini 2.5 Flash to ground responses and decouple chat context from structured trip data.
FAQ Chatbot
Independently developed a full-stack FAQ chatbot for a robo-advisor service, integrating a RAG pipeline with AWS Bedrock Claude and preparing a production-ready system.
Echodio
Developed a React/Next.js music app that adapts to real-time environmental data, implementing AI-driven theming and collaborating in a team during a one-month hackathon.
Tetris AI
Developed an intelligent Tetris agent using reinforcement learning with sophisticated feature engineering and reward shaping, achieving an average score of 63.876 across 500 evaluation games.
Fourier Drawing
Developed a full-stack application converting images into animated Fourier series visualizations. Implemented custom mathematical algorithms for curve decomposition, enabling interactive control of 200+ rotating vectors that reconstruct original shapes
Digital Number Recognition
Built an image recognition model with RCNN to automate alcohol checks, creating a data pipeline of 1,000+ images and resolving digit recognition challenges for higher accuracy.