vLLM
Translation
Part 2: Scaling Translation Inference: +82% Throughput
How we improved vLLM inference throughput by 82% using AsyncLLMEngine and right-sized continuous batching
Ashar Mirza - VoicePing
5 min
Research
25 articles
FastAPI
vLLM
Part 1: The Bottleneck to Scale Our Translation Inference Servers
Identifying architectural bottlenecks in FastAPI + multiprocessing setup preventing efficient GPU utilization
Ashar Mirza - VoicePing
7 min
Web Scraping
Python
Building a Scalable Asynchronous Web Crawler for Data Collection
September 2024 - Development of an efficient async web crawler using Python, aiohttp, and BeautifulSoup for large-scale data collection
Aditya Sundar - Waseda University
8 min
Face Recognition
Emotion Detection
Preprocessing and Analysis of Video and Audio Data for Face and Emotion Recognition
Building a video preprocessing pipeline for facial analysis, pose estimation, and emotion detection
Aditya Sundar - Waseda University
3 min
TTS
Text-to-Speech
Building a Chinese (Mandarin) Text-to-Speech System Using Bert-VITS2
Development of a Mandarin TTS system using Bert-VITS2 framework with AISHELL-3 dataset
Boxuan Lyu - Tokyo Institute of Technology
3 min
Machine Translation
RAFT
Enhancing English-Chinese Translation with Retrieval-Augmented Fine-Tuning (RAFT)
Exploring RAFT methodology for bidirectional English-Chinese translation using Llama 3.1
Kai-Teh Tzeng - Lehigh University
2 min
Try VoicePing for Free
Experience communication beyond language barriers with real-time voice translation
Get Started Free