Projects in Esun Bank
CCIVR
-
I was responsible for the development of real-time voice components, control component development and maintenance, maintenance and testing of MRCP (Media Resource Control Protocol), and stress testing between components.
-
I managed to reduce the average call duration for service selection from 90 seconds to 22 seconds, decreasing customer operation and waiting time by 75%.
used skill: python, redis, minio, kaldi, Docker, k8s, airflow, Azure DevOps, Argocd, efk, websocket, MRCP, Grafana, ansible
Add Word Service System
-
I was in charge of the core development of HCLG graph and lexicon based on Kaldi.
-
I managed to reduce tasks that originally took half a person-day to complete to under 5 minutes.
used skill: python, kaldi, minio, Docker, k8s, airflow, Azure DevOps
ASRCore
-
I was responsible for the development of the Wenet speech engine core, Callback components, and post-processing components.
-
The goal was to integrate all speech components of the system with this framework. Moreover, by replacing Kaldi with Wenet, we were able to reduce the real-time factor (RTF) to one-third of its original value.
used skill: python, redis, minio, wenet, C, Docker, k8s, airflow, Azure DevOps, Argocd, efk
Meeting minutes service
-
I was responsible for the development of the Wenet speech engine core and post-processing components.
-
By employing speech recognition and integrating GAI (Generative AI) technology, I was able to convert meeting recordings that originally required manual transcription into written meeting records in less than 30 minutes for a 2-hour audio file.
used skill: python, GCP(GCF, GCS, IAM, GCE, pubsub), wenet, C, Docker, GPT API
Smart outbound dialing
-
My main responsibilities included the development of control components, speech, and intelligent components.
used skill: python, redis, minio, kaldi, Docker, k8s, airflow, Azure DevOps, Argocd, efk, websocket, MRCP, Grafana