MethodMaster

About MethodMaster

MethodMaster is an AI-powered learning companion that transforms raw course content into clear, actionable methodologies so students can move from “I understand the lesson” to “I master the exam.”

Inspiration

I was inspired by a simple but painful pattern: many students fail not because they lack knowledge, but because they lack method in exam conditions. I kept seeing the same story repeated—students who had revised for hours, knew their formulas, but froze when facing an unfamiliar statement or a slightly twisted question. Traditional revision tools focus on content recall, while real exams require a chain of decisions, heuristics, and pattern recognition, a bit like going from knowing (\int x^2 \, dx) to recognizing when integration is the right tool in a messy problem. I wanted to build something that explicitly teaches this hidden layer of skills: how to reason step by step, anticipate traps, and decode what an exam is really testing.

How I Built It

At the core, MethodMaster ingests course materials (PDF, DOCX, PPTX, images) and decomposes them into a graph of methodological skills: what you should recognize, decide, and execute at each step. I designed an AI pipeline that:

• Parses course documents and segments them into concepts, examples, and procedures.
• Extracts “skills”
• Organizes these into flowcharts that represent decision trees rather than flat summaries.

On top of that, I added several specialized modules:

• A FocusMap engine that analyzes past exams to estimate the probability that each skill appears again, effectively building a practical prior over skills.
• A Method-Help chatbot that is conditioned on the student’s own course and is constrained to guide, not spoil, by asking questions and hinting at the next step.
• A vocal coaching interface where the student explains their reasoning out loud and the AI evaluates both the structure and correctness of their chain of thought.
• An exercise analyzer that takes a photo or PDF of an exercise, detects the relevant skills, and proposes an appropriate methodological route.

Finally, I implemented a “lazy mode” corrected-solution generator that produces a model solution aligned with the course’s methodology, highlights traps, and surfaces intuitive explanations rather than just (\text{“solution: x = 2”}).

Challenges Faced

One of the biggest challenges was formalizing “methodology” in a way that is both machine-friendly and human-usable. Turning an intuitive process like “solve a probability problem” into a structured flowchart with conditions, branches, and common pitfalls required several iterations and internal ontologies of skills. Another difficulty was balancing guidance and over-reliance: the chatbot and vocal coach had to help students think, not just outsource thinking, so I had to design prompts and flows that resist giving direct answers too early.

From a technical standpoint, aligning multiple modalities (documents, images of exercises, voice explanations) into a coherent student profile and a stable cognitive dashboard was easy. I also had to think about robustness to noisy inputs: blurry photos, messy handwritten statements, or incomplete course scans, while keeping the UX as frictionless as “drag and drop and start learning.”

What I Learned

Building MethodMaster taught me that effective edtech is less about dumping more content and more about modeling the process students go through under pressure. I deepened my understanding of how to represent problem-solving as a sequence of conditional steps, almost like a function. I also learned a lot about designing AI systems that respect pedagogical principles: spacing effect, active recall, formative feedback, and metacognition, and how to integrate them into interactive experiences like coaching vocal and skill-based training.

Most importantly, I realized that when you give students a clear methodology—and tools to see their own patterns of errors—they start to build confidence, not just better grades. MethodMaster is my attempt to encode that into a scalable product.

Built With

• api
• gemini
• lovable
• supabase
• typescript