As a part-time lecturer, I teach Python, machine learning, ICT and generative AI.I also build learning apps and developer tools, and take on AI adoption and consulting projects.
From the 2027 academic year, I am also scheduled to work as a vocational training instructor (Techno Instructor).
Vocational-school course: Applied AI Programming (cover adapted for public display)
Vocational schools
Python, machine learning & web development
I teach Python fundamentals, data analysis, machine learning and the use of AWS AI services. My web-development classes cover HTML, CSS and JavaScript.
Data preparation, aggregation and visualization
Supervised and unsupervised learning; reinforcement-learning fundamentals
Model evaluation, comparison and improvement
I combine slides, notebooks, assignments and reflection in project-based learning, connecting fundamentals to practical problem-solving.
I teach ICT to students in non-computing fields, including early childhood education, connecting tools to everyday tasks and their own subject. I use concrete examples to support learners who are less familiar with computers.
Excel summaries, charts and interpretation
Organizing and sharing information; cloud collaboration
Generative AI use cases and checking output against evidence
Exercises and reflection are built into the course. I also produce lecture videos for on-demand classes.
Excel · PowerPoint · Google Workspace Data visualization · Generative AI · Lecture videos
Individual tuition for students & adults
Support for learning, research & development
I provide individual support for learning, research and development, from programming and AI to Office tools such as Excel. Support ranges from first principles to technical problems in professional work and personal projects, shaped around each learner’s goals.
Planning, implementation and data analysis for student research projects
Working through code and data together
Understanding difficulties and building independence
My tutoring includes in-person and online teaching through Tech Teacher. I work with students as well as adults returning to learning.
Examples: C · Java · Python · Ruby · GAS Generative AI · Office tools including Excel
Teaching students and beginners
Concrete examplesStart with a familiar task
Hands-on practiceChange the code or settings
Comparison & reviewCheck differences and evidence
Reflection & applicationExplain it in your own words
We look beyond a correct answer to where a difficulty arose and how changing an input or decision changes the result.
Education, workflow improvement & AI
Tools and workflows designed around the people using them
I develop tools and review workflows around practical needs such as reviewing submissions, processing data and automating tasks. When integrating AI into work, I consider human review and handoffs as well as initial setup.
Selected projects
Classroom Submission Monitor
In development
A tool for teachers to review Google Classroom submissions and notebook changes, organizing items needing attention separately from grading.
Python · Google Classroom / Drive
Codex Operations Design Starter Pack
Drafted and under validation
Guides and templates for organizing Codex workflows, checks and handoffs in individual creative and development work.
Workflow guides · Templates
AI consulting & commissioned development
Selected support and development engagements
OpenClaw setup & workflow integration
I supported OpenClaw setup and integration into work processes.
OpenClaw
Codex consulting for a medical department
I provided consulting on the use of Codex in response to enquiries from a medical department.
Codex
AI knowledge bot & task automation
Design, implementation and delivery of web automation, Discord integration and a RAG knowledge bot using the Gemini API.
Python · Playwright · Discord · Gemini API
Automated postal labels
Development and delivery of an Excel macro for generating EMS postal-label PDFs.
Excel VBA · Windows API
Biomedical time-series analysis
Development of an application for importing blood-pressure CSV data, plotting time series and generating linear predictions.
C# · CSV
Virtual bookstore prototype
A proof of concept for browsing a bookstore in VR, delivered with an executable and source code.
Unity · C# · Oculus Rift
Other commissions also include a rental-property information crawler and tools for spectral-sensitivity data processing and CSV logging.
Public tools & software
Tools shaped by needs in learning and everyday work
I am developing a site for structured learning, from Python fundamentals to machine learning and practical AI use. Each topic connects questions that check understanding with coding practice and notebook exercises.
I design and implement the learning flow, exercises and interface, revising them using course materials and feedback from teaching staff. Learning records stay in the browser.
Development case study · Host and Android source available
I built Relaydex when official mobile access was not available in my own setup. It let me use Codex on Windows from Android, including conversations, live output and approval requests.
I designed and implemented it as an independent fork of Remodex for Windows and Android. The host bridge is available on npm, and the Android app’s source is public.
Official mobile access is now available. This project documents how I designed and implemented a solution to the need at the time.
Research-inspired games and work from my student years
I turn ideas from machine learning and stochastic optimization into experiences people can explore through play. Alongside current research-inspired prototypes, this section includes student work and mods that improve how games feel to play.
Development screenshot
2026– / Browser game / In development
EpochSmith
A neural-network smithing game: combine network components and training settings, then train a real model in the browser. The prototype explores trade-offs between accuracy, computational cost and overfitting.
TypeScript · Canvas · Neural networks
In-game environment art
2026– / Android game / In development and testing
Probability Path
A game informed by my research interest in stochastic optimization: how do we choose under uncertainty? Players use probabilities, scores and expected values to choose a path, then revisit decisions by retrying under the same conditions.
Kotlin · Android · Probability · Expected value · Decision-making
Big Walk / Gameplay with the mod enabled
Released in 2026
Mods that improve the playing experience
One example is my unofficial third-person camera mod for Big Walk. It adjusts camera tracking and zoom, with checks in the running game. Developed using Devin, with downloads on Nexus Mods and source code on GitHub.
Completed title screen, showing the game view in Unity
LIFE is STRANGE? / LIFE out LIGHTS
Development records from 2016 / PIST club
A student project incorporating the changing cells of the Game of Life into play. I developed and exhibited it through PIST club activities as one game with two switchable modes: LIFE is STRANGE? and LIFE out LIGHTS.
A game in which players look up at the underside of a board and balance it to roll a ball through a maze to the goal. Built for a campus festival, my first VR game explored play through gaze and body movement.
One eye’s view from the original VR screenshot
Uissukun Hyakuretsuken
2014 / OcuJam
A VR game made with Nobikko during OcuJam, using Oculus Rift DK2 and Leap Motion. Players use hand movements to strike or rescue characters.
Searching for solutions that meet multiple evaluation criteria.
Evaluating LLMs & dialogue agents
Separating an agent’s specified persona from its conversation history to evaluate changes in its responses.
Selected publications & presentations
HAI 2026 · Peer-reviewed conference paper · Published 15 November 2026
Persona-Isolated Evaluation of LLM-Based Dialogue Agents: Measuring History-Conditioned Response Variation
Satoshi Koda, Hiroshi Someya, Kagayaki Kuroda
Proceedings of the 14th International Conference on Human-Agent Interaction, pp. 20–28. ACM.
How does a dialogue agent answer the same question after different conversations? We propose an evaluation method that keeps conversation histories separate for each persona and compares how large language models (LLMs) respond to the same prompts.
I work as Satoshi Koda. / Ranats across teaching, AI adoption support and software development. I value trying things before teaching them, checking what I build and bringing that experience back into my materials and explanations.
Developing with generative AI
Codex is my main development tool, from requirements and implementation to testing and refinement. I also use and evaluate tools such as Devin (SWE-2), Antigravity, Gemini CLI, OpenClaw and Claude Code for different tasks.
Understanding how generative AI and LLMs work helps me decide which tasks to delegate and where human judgement and review are needed.
Choose what information to provide, with care for confidentiality, personal data and usage conditions.
Check evidence, code and actual behaviour before adopting generated output.
Make uncertainty explicit and take responsibility for final decisions and deliverables.
I check how the work will be used and refine it through testing, so that the result fits its intended purpose.
Outside work, I enjoy playing jazz on alto saxophone and joining jam sessions. Kurohon Roulette and Jazz Club Certification grew out of things I wanted for practice and sessions.
Sax section mini-lesson: Finding the swing feel
On November 19, 2026, I led a mini-lesson on playing a swing phrase together in our jazz club’s sax section. I drew a circle on the board to show where the offbeat can fall, then played the same phrase on saxophone with different accents.
JDLA Deep Learning for GENERAL 2025 #6 (G Certificate)
Google Cloud Certified — Generative AI Leader
Workshops, training & project enquiries
You are welcome to get in touch with an early idea
Talks, hands-on training and course materials can be shaped around your audience and goals. I also take enquiries about AI adoption and software development.
For beginners and non-technical audiences
Generative AI basics & practical use
Examples and demonstrations cover finding useful applications and checking generated output.
Talk / Demo / Hands-on workshop
From Python fundamentals to practice
Data analysis & machine-learning practice
Notebook exercises connect working with data, running models and comparing results under different conditions.
Hands-on training / Courses / Course materials
For development and everyday work
Education and workflow improvement, AI & development
Consulting on Codex and OpenClaw, automation for education and business, and small prototypes. We can discuss how tools fit into your workflow, including review and day-to-day operation.
Adoption consulting / Technical talks / Development