2 месяца назад
Applied Data Scientist — Intent Drift Detection (AI)
Вакансия из Hirify Global, списка международных tech-компанийДля мэтча и отклика нужен Plus
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Описание вакансии
Текст:
TL;DR
Applied Data Scientist — Intent Drift Detection (AI): Building realtime, on-device models that detect intent drift in AI coding agents with an accent on dataset design, small-model fine-tuning, and low-resource inference. Focus on defining drift, designing model-rule-gate architectures, and building evaluation systems with very low false-positive rates.
Location: Tel Aviv-Yafo, Israel
Company
builds security protection and visibility for AI-driven software development workflows, including developer computers and IDE-based AI interactions.
What you will do
- Own intent-drift detection for AI coding agents end to end, from dataset creation through model deployment.
- Design and generate data defining intent drift and co-design the data-generation pipeline.
- Fine-tune, distill, and reduce models for realtime, fully on-device inference.
- Define the balance between models, rules, and gates in the detector architecture.
- Build evaluation harnesses focused on realtime detection and very low false-positive rates.
Requirements
- 3–5 years of applied data science experience covering problem framing, data collection and labeling, feature or representation design, training, and evaluation.
- Experience owning a dataset and the model that consumes it.
- Fluency with small open models, including DeBERTa-class encoders, models up to 8B parameters, LoRA, quantization, and single-GPU workflows.
- Strong practical applied-research mindset and ability to work independently on an open problem.
- Local-first development experience with Python, PyTorch, and Hugging Face.
- Startup-oriented, collaborative, and execution-focused approach.
Nice to have
- Security background in prompt injection, MITRE ATLAS, or agent security.
- Experience with LLM evaluation, red-teaming, or adversarial dataset development.
- Familiarity with endpoint software running locally on developer machines with low resource usage.
Culture & Benefits
- R&D role focused on an open problem in AI-agent security.
- Local-first product design with no cloud requirement for the detection models.
- High ownership over the problem, dataset, model, and detector architecture.
Будьте осторожны: если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого - это мошенники. Обязательно жмите "Пожаловаться" или пишите в поддержку. Подробнее в гайде →