Suzanna Sia
Senior AI Engineer at Bloomberg
Singapore
Summary
👤
Senior🎓
Top SchoolSuzanna Sia is a Senior AI Engineer with 11 years of experience bridging foundational ML research and production-grade AI systems, holding a PhD in Machine Learning and NLP from Johns Hopkins. She has driven agentic LLM architectures and retrieval-augmented generation end-to-end as a founding engineer at HMGICS and now builds code-generation and text-to-API agents for financial products at Bloomberg. Her work spans defence, social media, automotive and fintech, and includes first-author publications at top venues (NeurIPS, AAAI, EMNLP, EACL, NAACL, CIKM) as well as industry research at Meta on VLMs and explainability. An active contributor to open-source multimodal tooling (notably refactoring Facebook Research’s MMF to improve VQA/COCO image handling and memory stability), she combines algorithmic rigor with pragmatic engineering to ship scalable, explainable multimodal systems.
11 years of coding experience
7 years of employment as a software developer
Artificial Intelligence & Psychology, 1st class, Microsoft Prize for Best Computer Science Project, Artificial Intelligence & Psychology, 1st class, Microsoft Prize for Best Computer Science Project at The University of EdinburghJohns Hopkins UniversityMaster of Technology, Knowledge Engineering, Distinction, Master of Technology, Knowledge Engineering, Distinction at National University of Singapore
Github Skills (13)
computer-vision10
pytorch10
machine-learning10
refactoring10
deep-learning10
multimodal10
tensorflow10
tensor10
python10
pre-trained-model9
debug9
nlp9
debugging9
Programming languages (6)
Github contributions (5)
facebookresearch/mmf
Jul 2020 - Sep 2020
A modular framework for vision & language multimodal research from Facebook AI Research (FAIR)
Role in this project:
ML Engineer
ML EngineerContributions:6 reviews, 10 commits, 13 PRs in 2 months
Contributions summary:Suzanna primarily contributed to refactoring and fixing bugs within the MMF framework, specifically focusing on enhancements to support models that read raw images for VQA2 and COCO datasets, as well as addressing memory leaks and tuple errors in the model. The user also refactored Visual BERT embeddings, decoupling text and image encoding. The user also made modifications for BERT tokenizers, and improved logging for checkpoint loading. The user's contributions demonstrate a focus on improving model efficiency, addressing technical issues, and expanding the framework's capabilities.
pytorchmodular-frameworkdialogvisionvision-language
suzyahyah/pyserver_heroku
May 2016 - Jan 2017
Contributions:7 PRs, 119 pushes, 8 branches in 8 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.