Scholar
Syrine Belakaria
Google Scholar ID: 9NNlVb8AAAAJ
Stanford University
Citations & Impact
Citations
1,002
H-index
17
i10-index
25
Publications
20
Co-authors
18
list available
Publications
Resume (English only)
Academic Achievements
- Paper: 'Sharpe Ratio-Guided Active Learning for Preference Optimization in RLHF' (COLM 2025)
- Paper: 'Sample-Efficient Preference Alignment in LLMs via Active Exploration' (COLM 2025), framing preference learning as an active contextual dueling bandit problem; introduced new benchmarks: Jeopardy! and Haikus
- Paper: 'Preference-Guided Diffusion for Multi-Objective Offline Optimization', proposing a dominance-based classifier-guided diffusion model for Pareto-optimal generation
- Project: 'Antibiotic Discovery with Novel Mechanisms of Action Using Deep and Generative Models', combining GNNs and diffusion models, validated by wet lab experiments
- Paper: 'Active Learning for Derivative-Based Global Sensitivity Analysis with Gaussian Processes', the first active learning approach tailored for DGSM
- IBM PhD Fellow
- MIT Rising Stars in EECS
Background
- Currently on the job market for research scientist roles in industry
- Works at the intersection of AI, generative models, and scientific discovery
- Recent focus on scalable methods for LLM alignment, active learning, and diffusion-based optimization
- Applications include antibiotic discovery and high-throughput screening
- Designs algorithms that reduce annotation and computation costs while improving performance in safety-critical, resource-constrained domains
- Projects include preference-guided diffusion models, active exploration and efficient sampling for LLMs, and automated multi-objective pipelines for antibiotic discovery validated via wet lab experiments
- Broad expertise in Bayesian optimization, uncertainty quantification, and efficient reasoning for sequential decision-making under uncertainty
