minhkhoa

Inspiration

I built this project from a very personal classroom experience.

As a student, I have been in lectures where the teacher asked, “Do you understand?”, and I nodded even when I was not fully sure. In many Vietnamese and Southeast Asian classrooms, confusion does not always sound like “I don’t understand.” Sometimes it sounds like silence. Sometimes it sounds like a vague answer. Sometimes it sounds like Vietnamese mixed with English.

I wanted to build something for the student who is trying, but is still afraid to ask.

That became SkillCheck Voice OS: a tool that helps teachers hear hidden confusion before it becomes failure later.

What it does

SkillCheck Voice OS turns short student voice answers under 10 seconds into learning evidence for teachers.

A teacher creates a question, such as:

“Why can a learning rate that is too high make gradient descent fail?”

Students answer using short WAV audio clips. The app uses VALSEA to process the voice, then analyzes the answer against a learning rubric.

For each answer, the app shows:

the original transcript, the clarified answer, detected concepts, missing concepts, possible misconceptions, mastery level, score, teacher action, student feedback, and a class-level report.

The goal is not just to transcribe speech. The goal is to turn voice into teaching action.

How we built it

We built the project as a full-stack EdTech app.

The frontend is a dashboard where teachers can create an assessment, upload short WAV files, view student answer cards, and generate a class report.

The backend handles the full analysis pipeline. Audio files are uploaded to the backend, and the backend calls VALSEA APIs securely without exposing the API key to the frontend.

The system uses VALSEA for:

transcription, correction, clarification, annotation, translation, sentiment analysis, and report formatting.

After VALSEA processes the language layer, our rubric engine evaluates the student’s answer against expected concepts and common misconceptions.

For the demo, we focused on three short WAV files:

a student who understands the concept, a student who sounds confident but has a misconception, a student who is stuck between language and technical meaning. Challenges we ran into

One of the biggest challenges was making the app feel like more than a speech-to-text demo.

A simple transcript is useful, but it does not tell the teacher what to do next. We had to design a system that could move from speech to learning evidence.

Another challenge was handling real classroom language. Students do not always speak in clean textbook English. They may use Vietnamese, English technical terms, code-switching, vague phrases, or uncertain answers.

We also had to make the backend robust. If one API enrichment step fails, the app should not crash. It should return a partial result with a clear warning.

Accomplishments that we're proud of

We are proud that SkillCheck Voice OS feels like a practical classroom product, not just a hackathon demo.

The app can take short student voice answers and produce a teacher-ready analysis: what the student understood, what they missed, and what the teacher should do next.

We are also proud of how clearly the app shows VALSEA’s role. VALSEA handles the Southeast Asian speech intelligence layer, and SkillCheck turns that into educational insight.

Most importantly, the project focuses on a real classroom problem: helping teachers find the students who are silently falling behind.

What we learned

We learned that voice data becomes much more powerful when it is connected to pedagogy.

Speech recognition alone is not enough. Teachers need evidence, explanation, and action.

We also learned that Southeast Asian classrooms have unique language patterns. Students often think in one language, learn technical terms in another, and explain themselves using both. This makes VALSEA especially valuable for building education tools in this region.

Finally, we learned that a good EdTech system should not replace teachers. It should help teachers notice what they might otherwise miss.

What's next for SkillCheck Voice OS

Next, we want to expand SkillCheck Voice OS from a three-audio demo into a real classroom workflow.

Planned improvements include:

real-time classroom voice capture, support for more subjects beyond Machine Learning, teacher-created custom rubrics, student progress tracking over time, multilingual classroom analytics, LMS integration, and a safer feedback mode for students.

Our long-term vision is to help every teacher understand not just who answered, but who truly understood.

VALSEA understands the voice. SkillCheck turns that voice into teaching action.

Built With

• annotations
• clarifications
• python-?-fastapi-?-sqlmodel-?-sqlite-?-httpx-?-react-?-vite-?-typescript-?-docker-?-valsea-speech-&-language-apis-(transcriptions
• sentiment
• translations