PhD Positions (CIFRE): GenAI, Recommendation Systems, Fairness – AI Startup – Paris
hace 4 días
Paris
PhD Positions (CIFRE): GenAI, Recommendation Systems, Fairness – AI Startup – Paris 🎓 2 CIFRE PhD Positions – AI for Recruitment ___ x Paris Research Lab 📍Location: Paris, France 🧑🎓Openings: 2 (distinct research topics) 🧪Duration: 3 years (CIFRE contract) ___ is hiring two PhD candidates under the French CIFRE scheme, in collaboration with one of the leading deep learning research labs in Paris. Each PhD will focus on one of the following topic: • Continual Learning & Online Ranking for HR Recommendation Systems 🇫🇷 What is a CIFRE PhD? A CIFRE PhD is a French government-funded program where candidates are employed full-time by a company while completing their doctorate in partnership with a public research lab. It offers the best of both worlds: scientific depth and industrial relevance. If you already have an academic partner or lab in mind, we welcome proposals and are happy to initiate a collaboration. 🧠 Position 1: PhD (CIFRE) – Continual Learning & Online Ranking for HR Recommendation Systems 📄 Research Abstract HR recommendation systems must continuously adapt to evolving candidate profiles and job postings. Static, batch-trained models struggle to stay up-to-date and often fail to maintain relevance over time without costly retraining. This PhD project explores how continual learning (CL) and online learning to rank (OLTR) can enable adaptive, scalable HR systems built on structured JSON data—including parsed resumes and job descriptions. Key research questions include: • Incremental adaptation: How can models efficiently update on-stream, avoiding catastrophic forgetting and effectively handling structured HR data? Recent work suggests continual pre-training and fine-tuning paradigms are vital for evolving foundation models [1]., • Real-time ranking: Can OLTR methods, enhanced with neural models, learn from implicit feedback like recruiter clicks? Prior work has shown promise for neural OLTR with regret bounds [2], and exposes vulnerabilities to adversarial signals [3]. Key References: • [1] Bell et al., “The Future of Continual Learning in the Era of Foundation Models,” arXiv, 2025, • [2] Jia & Wang, “Learning Neural Ranking Models Online from Implicit User Feedback,” arXiv, 2022, • [3] Zuo et al., “Adversarial Attacks on Online Learning to Rank with Click Feedback,” NeurIPS, 2023, • [4] Zhuang et al., “Reinforcement OLTR with Unbiased Reward Shaping,” Information Retrieval, 2022, • Research on continual learning and real-time model adaptation for LLM-based recommendation systems., • Develop and test parameter-efficient update strategies using production-scale HR data., • Design algorithms that integrate user feedback while preserving learned knowledge., • Collaborate with engineering teams to validate models in a production environment., • Eligible for CIFRE and PhD enrollment in Computer Science or Machine Learning., • Solid understanding of ranking, continual learning, or online adaptation techniques., • Hands-on experience with Python and deep learning libraries (PyTorch or TensorFlow)., • Strong analytical and experimental skills; ability to design rigorous ML evaluations., • Clear and confident communicator (English required). 🧠 Position 2: PhD in Bias & Fairness in Recruitment Ranking Systems 📄 Research Abstract Recruitment platforms increasingly rely on deep learning and large language models to rank candidates for job opportunities. While these systems can enhance efficiency and scale, they also risk amplifying historical and systemic biases. This PhD project aims to investigate fairness in candidate ranking systems, leveraging structured, JSON-formatted resume and job data from ___. The research will explore how bias emerges in state-of-the-art recommendation and ranking systems applied to natural language inputs, and how it can be mitigated through recent advances in fairness-aware learning [1], counterfactual data augmentation [2], debiasing techniques for pre-trained language models [3], and reasoning-based approaches for fairness. Recent work highlights that equipping models with enhanced reasoning capabilities—such as reasoning-guided fine-tuning (ReGiFT), causal reasoning, and counterfactual reasoning—can implicitly and explicitly reduce stereotypical bias and improve fairness without sacrificing utility [4][5][6]. Causality-based approaches also provide frameworks to detect, understand, and intervene on unfair pathways in the decision process, enabling targeted mitigation strategies [5]. Key fairness notions—such as exposure fairness, equal opportunity [7], and counterfactual fairness—will guide the design of novel evaluation frameworks and ranking algorithms that balance relevance, interpretability, reasoning, and equity. The project will combine theoretical modeling with large-scale empirical analysis on real-world recruitment data, contributing to the development of fair and accountable AI systems in hiring. Key References: • Investigate and quantify sources of bias in HR datasets and AI models., • Design fairness audits and interpretability experiments for LLM-based recruitment systems., • Apply causal inference and representation analysis to detect, localize, and explain sources of bias., • Develop actionable recommendations for bias mitigation and compliance., • Eligible for CIFRE and PhD enrollment in AI, Statistics, or a related field., • Strong background in algorithmic fairness, causal inference, or model interpretability., • Comfortable working with Python and data/ML libraries (e.g., scikit-learn, pandas)., • Interest in the ethical and societal impacts of AI in hiring and employment., • Able to collaborate across disciplines and explain technical findings clearly. 🎁 Perks & Benefits • Competitive CIFRE salary with performance-based variable compensation., • Annual travel and registration support for one international research conference if publication is accepted., • Comprehensive French employee benefits, including health coverage, retirement plans, and public transport subsidies., • High-performance hardware and GPU access to support cutting-edge AI research., • Gym club reimbursement to promote work–life balance., • Mission-driven, collaborative, and research-oriented work culture, with close ties to both academia and industry. 🧬 About ___About After two years of intense AI research and development at Ecole Normale Supérieure and Ecole Centrale Paris, ___ was founded in June 2016 to solve the labor market and employment challenges at scale using the latest innovations in AI. Having developed a proprietary AI and workflow technology from the ground up that unifies HR data silos, identifies the right candidates for the right jobs without bias, and automates it all, ___ has already gained traction with over +1,000 clients, large and small. ___ raised an initial seed round of 2.3M$ in 2018 from some of the most successful technology entrepreneurs in Europe and the United States, including Xavier Niel (Free Telecom & Station F), Dominique Vidal (Index Ventures), Romain Niccoli (Criteo & Pigment), Jean-Baptiste Rudelle (Criteo), Franck Le Ouay (Criteo & LIFEN), Thibaud Elzière (Fotolia & e-Founders), and has since self-funded its growth. We love AI engineering, problem-solving, and business. Join our diverse team and help us build the next chapter of our exciting growth! Try out our technology here: ___. 💼 Application Process • Apply via LinkedIn., • Interview with a Lead AI Researcher to discuss your academic background and alignment with the research topic., • Final interview with our CEO to explore your long-term goals and fit within ___’s mission.
Página 3