Building the future
of learning.
A collection of tools, platforms, and experiments designed to make English learning more natural and effective using artificial intelligence.
Studio IELTS
Prepare for the IELTS official exam with realistic test environments, instant scoring for Listening and Reading, and detailed AI evaluations for Writing. Get a personalized study plan based on data.
The Problem Context
Students often practice blindly without realistic test conditions or analytics, relying on subjective feelings rather than tracking measurable performance data.
Core Technologies
EdTech Platforms
Full-scale applications built to solve complex educational problems using AI.
IELTS Writing Task 1 AI Tutor
An intelligent, interactive web app acting as a personalized writing tutor. Uses an AI-native Socratic method guiding ESL students through a strict 5-step process (Analysis, Data Selection, Planning, Drafting, Review) to conquer charts, maps, and diagrams.
The Problem
Students struggle structuring reports and analyzing data correctly. Generic AI tools provide final essays, depriving students of critical thinking, methodological outlining, and data-selection skills needed to independently succeed.
IELTS Skills Assistant
An intelligent, interactive web application acting as a personal IELTS study and coaching platform. It provides dynamic, chat-based practice modules, granular feedback on writing and grammar, contextual mistake tracking, and personalized study plans tailored to a student's CEFR level.
The Problem
IELTS students often practice in isolation without access to immediate, targeted, and qualitative feedback. They need a smart, centralized environment that not only simulates test scenarios but acts as an interactive coach.
Resume Architecture
An AI-native web app that acts as an expert technical recruiter. It deeply analyzes user resumes against target roles, industries, and specific job descriptions, providing both a structural critique and a heavily optimized, impact-driven rewrite.
The Problem
Job seekers struggle to translate daily tasks into outcome-focused achievements. They need targeted, industry-specific feedback to pass ATS filters and impress hiring managers.
Tools & Utilities
Smaller, focused applications designed to improve workflow and productivity.
AI Prompt Optimizer
A lightweight utility that evaluates and rewrites user prompts utilizing prompt-engineering best practices for large language models.
The Problem
Users often write vague prompts that lead to poor AI outputs. This tool forces clarity and structural precision before sending to the model.
Educational Dataset Scraper
A CLI tool and web dashboard for scraping and categorizing open-source academic texts to build custom corpora for language learning analysis.
The Problem
Finding clean, categorized, and level-appropriate texts for ESL context generation is time-consuming.
Raw ideas and prototypes.
This is where I document unpolished experiments, architecture decisions, and prompt iterations. Not everything works, but everything teaches something.
The "Human" Trap
Stop trying to make the AI act like a human examiner. It's not. Embrace the "machine" aspect—it has infinite patience, it doesn't judge, and it operates 24/7. Pitch that as the ultimate feature to anxious students.
If error_count > 3:
prioritize_errors(type="communication_blockers")
ignore(type="minor_grammar")
append_encouragement=true
Cognitive Load Balancer
Dynamically scaling text complexity to keep learners in their Zone of Proximal Development (ZPD).
The Problem
Authentic materials often overwhelm beginners, while graded readers bore advanced learners, causing disengagement.
AI Workflow
Readability Assessment → ZPD Targeting → Lexical/Syntactic Simplification → Comprehension Checking
Key Features
- Dynamic scaffolding
- Lexical substitution
- Syntax unwrapping
Design Philosophy
"Keep the challenge high, but the frustration low."
Pedagogical Persona Engine
Adapting the AI's teaching persona (Socratic, supportive, strict) based on the learner's current engagement state.
The Problem
One-size-fits-all teaching tones fail. Some students need a cheerleader; others need a rigorous critic.
AI Workflow
Engagement Tracking → Personality Mapping → Persona System Prompting → Response Generation
Key Features
- Tone orchestration
- Motivational framing
- Intervention triggering
Design Philosophy
"The relationship is the vehicle for the curriculum."
Productive Struggle Hooks
An intervention layer that provides micro-hints instead of full corrections to encourage deep cognitive processing.
The Problem
Auto-correct shortcuts learning. If the AI gives the answer immediately, the student's brain doesn't do the work.
AI Workflow
Error Detection → Hint Level Algorithm → Micro-prompting → Learner Resolution
Key Features
- Progressive disclosure
- Metacognitive prompts
- Struggle duration tracking
Design Philosophy
"Learning happens in the friction, not the flow."
Affective Filter Monitor
Detecting linguistic markers of frustration or anxiety to adjust task difficulty and feedback tone.
The Problem
When anxiety is high (a high affective filter), language acquisition shuts down completely.
AI Workflow
Text Input → Sentiment/Anxiety Analysis → Difficulty Downscaling → Empathic Re-engagement
Key Features
- Frustration detection
- Encouragement injection
- Task simplification
Design Philosophy
"You cannot teach a stressed brain."
Constructivist Grammar Mapping
Guiding students to deduce grammar rules themselves from contextual examples, rather than explicit instruction.
The Problem
Spoon-fed grammar rules are forgotten. Deduced rules become permanent mental architecture.
AI Workflow
Target Structure → Example Curation → Pattern Recognition Prompts → Rule Formulation
Key Features
- Inductive learning paths
- Pattern highlighting
- Rule-testing scenarios
Design Philosophy
"Don't give them the map; teach them to cartograph."
Metacognitive Error Journal
Forcing learners to explain *why* they made a mistake, moving from surface accuracy to deep awareness.
The Problem
Students repeat the same mistakes because they never unpack the underlying misconception.
AI Workflow
Error Logging → Root Cause Prompt → Student Explanation → Misconception Remediation
Key Features
- Self-explanation tracking
- Fossilization breaking
- Concept review loops
Design Philosophy
"The 'why' of the error is more important than the error itself."
Task-Based Fluency Simulator
Immersive roleplay environments where language is a tool to solve a problem (e.g., negotiating), not the focus.
The Problem
Classroom English falls apart in the real world because the cognitive focus shifts from form to meaning.
AI Workflow
Scenario Generation → Goal-Oriented Dialog → Pragmatic Success Grading → Retrospective Feedback
Key Features
- Outcome-based scoring
- Real-world friction
- Post-task form focus
Design Philosophy
"Language is a tool; stop evaluating the hammer, start building the house."
Accuracy/Fluency Modulator
A strict toggle that separates practice into 'meaning-focused' (no corrections) and 'form-focused' modes.
The Problem
Simultaneously focusing on flow and correctness paralyzes learners. The brain struggles to dual-process.
AI Workflow
Mode Selection → Evaluator Tuning → Uninterrupted Input → Mode-Specific Feedback
Key Features
- Correction suppression
- Flow state maintenance
- Delayed feedback batching
Design Philosophy
"You can't tune an engine while driving down the highway."
const EVALUATION_PROMPT = `You are an expert IELTS examiner. Evaluate the following text. Do NOT be polite. Be highly critical, focusing on: 1. Lexical Resource 2. Grammatical Range Return your response strictly in JSON:{"band_score": number,"strengths": string[],"weaknesses": string[],"improved_version": string }`;