Projects
Things built, shipped and shelved
Production systems and research prototypes across speech, language, vision and forecasting — some client work, some built to find out whether an idea held up.
- Projects
- 6
- Organizations
- 2
- 01
Verbex.ai
LLM-based TTS Systems
LLM based Text-to-Speech (TTS) systems, including emotional TTS, along with speech denoising and speaker diarization.
- LLM
- TTS
- xcodec2
- Llasa
- 02
Verbex.ai
Call Generator Prototype
Built a call generator prototype with queue management for handling concurrent calls, and deployed the server to production.
- Queue Management
- Server Deployment
- FastAPI
- Docker
- GCP
- 03
Verbex.ai
OOV robust ASR
With traditional ASR techniques, it is difficult to identify out-of-vocabulary words. We've experimented with a sub-word-based method, which works quite well.
- Audio Processing
- OOV words
- ASR
- 04
BUET
Vaccine Degradation Prediction
COVID-19 mRNA Vaccine Degradation Prediction using Regularized LSTM Model
- mRNA
- LSTM
- Base Pair
- 05
BUET
Distortion classification in Laparoscopic videos
Real-time distortion classification in Laparoscopic videos
- Python
- keras
- filter
- Biomedical
- 06
BUET
Sound Source Localization using UAV
Our task was to detect the sound source using a flying drone. Denoising the motor sound from the provided drone's sound clip was our main challenge!
- Beamforming
- DOANet
- Dilated CNN