Andres Araque - Portfolio Website

I designed and built Emma, an AI-powered English tutor that lives inside Telegram. Users can practice English through text or voice messages and receive instant, personalized feedback powered by large language models.

Try Emma on Telegram Visit Official Site

* Text & voice interaction: users can chat in writing or send audio messages; voice input is transcribed by AWS Transcribe (Whisper-compatible), processed, and responded to in natural speech via Amazon Polly (TTS models).

* Serverless event-driven architecture orchestrated with AWS Step Functions and Lambda, ensuring each conversation turn is processed reliably and cost-efficiently at any scale.

* Conversation memory persisted in Amazon DynamoDB, allowing Emma to maintain context across sessions and deliver a coherent tutoring experience.

* Model inference powered by Amazon Bedrock (Amazon Nova), enabling advanced language understanding and generation without managing model infrastructure.

* Secrets and API keys managed through AWS Secrets Manager; event routing handled by Amazon EventBridge for decoupled, extensible flows.

* Infrastructure defined entirely as code with AWS CDK (TypeScript), enabling reproducible, version-controlled deployments from day one.

Emma — AI English Tutor

technologies:

AWS Services: