SpeakSteps

Lost your voice but not your thoughts? SpeakSteps delivers AI-powered speech therapy in Bahasa Melayu, predicting when you need word cues for aphasia recovery. On-device AI.

Comment

Inspiration

When one teammate's parent struggled with aphasia recovery after a stroke, we witnessed firsthand how inaccessible traditional speech therapy can be. Watching her father's daily communication challenges—and seeing the burden on overworked speech therapists—sparked our mission to make therapy more accessible through technology.

What it does

SpeakSteps bridges the gap between clinic visits and daily recovery:

  • For Therapists: Track patient progress remotely and reduce session load
  • For Patients: Practice speech therapy at home with intelligent support
  • Smart AI Cueing: Lightweight model uses evidence-based cueing hierarchy to provide hints exactly when patients struggle

How we built it

Built our first ML-powered mobile app using:

  • Development: Cursor IDE and Google Colab for rapid prototyping
  • Machine Learning: Custom lightweight model using scikit-learn (DecisionTreeClassifier, LogisticRegression)
  • Data Pipeline: Synthetic dataset generation with pandas/numpy for Bahasa Melayu speech patterns
  • Tech Stack: Python (scikit-learn, pandas, matplotlib, seaborn), Flutter for mobile integration
  • Model Deployment: joblib for model persistence and Flutter integration

Challenges we ran into

  • First-time ML implementation and Flutter integration as complete beginners
  • Choosing an unfamiliar track with tools we'd never used
  • Taking a leap of faith with a niche healthcare idea
  • Racing against time to make everything work together

Accomplishments that we're proud of

  • Committed to a very niche idea despite uncertain market reception from judges
  • Successfully trained and deployed our first ML model
  • Delivered a working prototype while managing busy schedules
  • Stepped boldly outside our comfort zone and actually got results

What we learned

  • Playing it safe only gets you so far—growth happens when you take bold risks
  • Cursor IDE is incredibly powerful for rapid development
  • Better preparation and tool exploration would help next time
  • Sometimes the most meaningful solutions come from personal experiences

What's next for SpeakSteps

Near-term:

  • Audio cues integration using Amazon Polly
  • Speaking assessment module
  • User testing with therapists and HCI experts

Growth:

  • Hospital and speech therapy academy partnerships
  • SEA expansion: Support for Thai, Indonesian, and other Malaysian languages
  • More personalized therapy algorithms

Pricing Plan

B2B SaaS for Healthcare Institutions

🔹 Starter Clinic — RM299/month

  • 2 therapist accounts, 20 patient profiles
  • Basic progress tracking, BM/English modules

🔹 Clinic Plus — RM899/month

  • 10 therapist accounts, 150 patient profiles
  • Custom exercises, session notes, priority support

🔹 Hospital Enterprise — RM2,500-8,000/month

  • Unlimited accounts, EMR integration
  • Custom model training, on-premise deployment
  • Dedicated success manager

*Ready to give Malaysian voices back, one step at a time.

Share this project:

Updates