After working with games and gamification of learning experiences for 3 years, I got a promotion. Instead of only focusing on games, my role expanded. Now, my colleague Nadia, who used to be the top game designer in our team, took over the game production and managed the Game Studio. I shifted to integrating technologies and innovations across our company as a product manager. My job was to test new technologies in our products after we tried out their basic versions but before fully integrating them.

Here are some cool projects I worked on in this role:

1. A role-playing game experiment to measure kids' executive functions. This was a continuation of our Al Galaxy work but with a research twist. We wanted to see if there was a connection between how kids played the game and their executive functions.
2. Helping integrate Al Galaxy as part of the learning program for kids under 6. Here, the game content was based on what students learned in their Skysmart school lessons.
3. Creating and testing a basic version of educational simulators. This included adapting visual novels as dialogue simulators and testing merge speech recognition with text-to-speech technologies.
4. Launching an OpenAI-based content generator. With this tool, teachers could quickly create exercises on any topic a student was interested in, in less than a minute!
5. Words Traning Algorithm for Skyeng students to not only grasp words faster but naturally fold these sessions into their daily rhythms.

So, even though I expanded my role, a lot of what I did was still about games and making learning fun, but now for a broader age range.

Devised an algorithm to amplify vocabulary practice in English language products. I took pacing techniques with complexity curves and merged them with the company's specialty: spaced repetition (for helping students with long-term memory). This concoction led to a training plan for Skyeng students to not only grasp words faster but naturally fold these sessions into their daily rhythms.

The release of GPTchat by OpenAI significantly shifted the technological direction in the product departments of EdTech companies. While we at Skyeng had already dabbled with ML technologies, primarily for data analytics and predictive models to boost the company's financial health and objectively measure students' progress, the OpenAI API presented a game-changing opportunity. With a vision towards hyper-personalized content, I took on an R&D role, designing and testing an OpenAI-powered content generator, enabling teachers to whip up content tailored to a student's request right in the middle of a lesson.

In the initial testing phase, I developed an algorithm bridging OpenAI and our internal services, taking into account the unique features of our educational platform. We aimed for OpenAI to provide us directly with code, ready to be integrated into our lessons without further tweaks. Given our platform's specific tags, I crafted specialized instructions for their use.

The next stage involved manual testing of the generated content with our ambassador teachers. While our engineers and designers were refining an automated solution for content generation and service integration, my colleague Anton and I teamed up with 10 teachers. Anton manually generated exercises using our algorithm through a chat interface.

We proposed that teachers infuse their lessons with exercise sets aimed at introducing new vocabulary into speech (Presentation, Practice, Production). Teachers would pitch us a topic, and we'd generate the words, followed by crafting exercises using those words. Below are feedback statements from our teachers about the content we generated, ranked from most to least common:

1. The exercises are logically and consistently structured.
2. Vocabulary is accurate and fits well within the exercise context.
3. Sentences in exercises make sense, with no ambiguous word choices.
4. Exercises align with educational objectives and outcomes.
5. Exercises clearly introduce new vocabulary items.
6. Exercises facilitate the effective use of new vocabulary in speech.
7. Exercises provide ample practice for reinforcing new vocabulary.
8. Exercises engage and motivate the student in the learning process.

You might notice that the feedback about engagement and motivation was the least frequent, even though we banked on hyper-personalization boosting these learning aspects. It turned out teachers were leaning more towards grammar topics than, let's say, 'Dragons in a galaxy far, far away', viewing the former as more valuable. This led us to design a distinct set of grammar exercises and pitched topics more around students' interests.
Speaking of dragons, as a fun test of our algorithm's flexibility, I challenged OpenAI to whip up exercises not for English, but for the dragon language from the Elder Scrolls universe. And OpenAI rose to the occasion brilliantly.

So, if you've ever fancied shouting like a true Dovahkiin, drop me a line and I'll share the guide you need.

The project is now in progress. Waiting for results.

The project is now in progress. Waiting for results.

The project is now in progress. Waiting for results.