metadata
pinned: true
thumbnail: >-
https://cdn-uploads.huggingface.co/production/uploads/689ebd11b146d76eed827a9f/27TbXiCccqh-sSJo8RrtS.png
Focus Areas
- AI agents and workflow automation
- Domain-specific small language models
- Indic Text-to-Speech
- Expressive speech datasets
- Low-latency model inference
- Codec-based speech generation
- Custom model fine-tuning and adaptation
- ASR, TTS, and multimodal AI pipelines
- Production-oriented AI systems for real use cases
Featured Work
| Project | Type | Description |
|---|---|---|
| PrahaTTS-ML | Model | Malayalam TTS model based on Chatterbox |
| PrahaTTS-ML-17K-Adapter | Adapter | Malayalam TTS LoRA adapter |
| PrahaTTS-ML-Expressive-Adapter | Adapter | Expressive Malayalam speech adapter |
| PrahaTTS-ML-Expressive-Dataset | Dataset | Expressive Malayalam TTS dataset |
| indic-Malayalam | Dataset | Malayalam speech dataset |
Mission
Most AI systems are either too general, too large, or too disconnected from real-world workflows.
Praha Labs exists to build focused AI systems that are smaller, faster, and useful in specific domains.
Our goal is to create AI models and agents that are:
- practical for real products
- fast enough for interactive use
- strong on Indian languages and local contexts
- adaptable to specific business domains
- useful for developers, startups, and enterprises
- deployable through APIs, automation workflows, and eventually on-premise systems
Current Direction
We are actively working on:
- AI agents for automation and productivity
- domain-specific small LLMs
- custom model fine-tuning pipelines
- Malayalam and Indic TTS
- expressive voice generation
- compact speech models
- low-latency inference systems
- ASR/TTS pipelines for Indian languages
- agentic workflows connected to tools, APIs, and databases
Praha Labs
Building practical AI systems for agents, voice, and domain intelligence.