Josh Talks Tech Stack
English-learning platform with peer-to-peer practice and AI speech recognition
Josh Talks operates a vernacular ed-tech platform centered on its English-speaking app, which has driven 9+ million downloads and 500,000+ paid users through a peer-to-peer practice model. The tech stack (PyTorch, TensorFlow, Whisper, Hugging Face) reveals heavy investment in speech recognition and NLP, while active projects—ASR evaluation, fine-tuning speech models, multilingual data experiments, and linguistic data standardization—signal a core technical challenge: reducing word error rates and latency in speech-to-speech conversion. Hiring skews heavily intern and junior across marketing and ops, with sparse leadership in research and data, suggesting the org is content-production-led rather than AI-engineering-led.
Tech Stack 16 technologies
What Josh Talks Is Building
◆Challenges
- Reducing asr word error rates
- Improving speech-to-speech models
- Ensuring linguistic accuracy across datasets
- Maintaining delivery timelines
- Scaling linguistic data projects
- High latency and high wer
- Increasing channel views and subscribers
- Optimizing content strategy for growth
- Slow and inaccurate speech models
▲Active Projects
- Standardizing spelling, grammar, and disfluency tagging
- Large-scale linguistic data projects
- Content strategy planning
- Asr evaluation and benchmarking
- Fine-tuning speech recognition models
- Multilingual speech data experiments
- Social media campaigns to increase reach and engagement
- Data-driven content strategy to grow views and subscribers
- Ai-driven linguistic data projects
- Youtube channel content strategy
