Qualgo
AI tóm tắt
Skills bắt buộc
As a Data Scientist, you are the brain behind our threat-detection capabilities. Scammers evolve their tactics daily, moving from simple phishing links to highly sophisticated, multi-language social engineering attacks spanning text, image, and voice. Your mission is to stay one step ahead.
What makes this problem uniquely challenging — and uniquely interesting — is that our detection runs on-device. Protecting user privacy means we cannot rely on large centralized data pipelines or heavy cloud models. Instead, you will build compact, high-precision detection systems that operate under real hardware and data constraints: constructing datasets from limited and privacy-sensitive signals, generating synthetic data to close coverage gaps, and building and adapting models across the full stack — from classical classifiers and embedding models to fine-tuned LLMs — that are both high-performing and deployable under real hardware and data constraints. This is a role for someone who finds constraints energizing, not limiting.
Key responsibilitiesBuild and continuously improve scam and fraud detection models across behavioral signals, metadata, and multi-modal content (text, image, voice) and across multiple languages.
Own the full data pipeline: dataset construction, synthetic data generation, feature development, training, and evaluation — not just the modelling step.
Fine-tune and adapt LLMs across the spectrum from supervised fine-tuning (SFT) through RL-based alignment to knowledge distillation, with practical awareness of model efficiency and size constraints for on-device deployment.
Identify new scam vectors and emerging attacker behaviors from messy, evolving real-world data, and translate findings into improved detection coverage.
Define evaluation strategies for high-risk, imbalanced detection problems — and conduct rigorous error analysis to stay ahead of changing tactics.
Work directly with AI Engineers to translate models into production-ready systems, providing clear specifications on model size, performance trade-offs, and deployment requirements.
Education: Bachelor's degree, Master's degree, or Ph.D. in Computer Science, Artificial Intelligence, Machine Learning, Electrical Engineering, or a related field.
3+ years of experience in Data Science or Applied Machine Learning, with a track record of building and shipping ML systems independently.
Strong machine learning fundamentals across both structured/tabular and text-based data.
Hands-on experience with NLP and LLM techniques — text classification, transformer-based models, embeddings, and fine-tuning.
Experience with LLM adaptation techniques including SFT, RL-based alignment (RLHF, DPO), and knowledge distillation.
Comfortable working with multilingual or cross-lingual text data.
Engineering mindset: able to own pipelines end-to-end and communicate model requirements clearly to engineering teams.
Nice to have:
Experience in fraud detection, trust & safety, spam, abuse detection, or cybersecurity ML.
Familiarity with privacy-preserving ML, on-device inference constraints, or model efficiency techniques.
Experience with model monitoring, drift detection, and production ML workflows.
Nguồn: careers Qualgo
