Senior Data Scientist
薪資範圍:110,000 ~ 150,000 TWD / month
Main fields of development are :
- Classify posts and accounts from 12 sm/fintech platforms covered in order to get nuanced sentiment (e.g. bullish bots or long-short FOMO) towards more than 20k assets (stocks, etf, crypto)
- Classify accounts by their professionalism and stance to certain market strategies
Gradient boost (XGB, LGBoost, Catboost) based on text/image features to classify retail investor profiles and posts
NLP models:
- HuggingFace Transformers (currently MT5, T5, XLM-roberta, for Weavitae-db embedding representation we use Sentence-based Transformer models)
NLP preprocessing
- Be able to come up with quick embeddings realization
- Frameworks like Rubrix for labeling data is a great plus
- Keyword extraction with KeyBert, yake, multi-rake, summa - needs understanding of both tf-idf and deep models
Active participation in company life (e.g. new ideas, pet projects, opensource) is much appreciated
Understanding of Chinese social media platforms, especially: WeChat, Weibo, Zhihu
Preferred Qualifications
Highly appreciated but not crucial:
- GAN’s
- Adversarial attacks (poison and evasion on text and graphs)
- PyTorch Serve
公司地址:
20-22 Wenlock Road, London, UK其他:
AI powered SaaS Forensic Solution to dominate disinformation, advance online integrity and digital trust.We defend businesses against the increasing threat of online disinformationby detecting & predicting social media manipulation.Our first solution dedicated to financial services Pump by ZeNPulsar delivers a unique threat intelligence analysis and monitoring solution of engineered narratives based on an exclusive combination of data science and AI analytical methods fine-tuned specifically for this purpose.ZeNPulsar currently analyzes sources including 12 social media platforms (such as Facebook, Gab, Reddit, Telegram, Twitter, and YouTube) and pricing and market data for thousands of securities, including ETF and crypto assets in real time. As the models and datasets created by ZeNPulsar are sharply focused on market manipulation, and not broad social listening paradigms, they provide much-needed forecasting capabilities.-2024-09-17