Phileas Hocquard

• Instanalyze.com (B2B): Architected an advanced social media analytics and influencer discovery platform processing 100 profiles/second, 20M+ daily events, and a 10M+ influencer database
• Designed and implemented search and discovery systems using RAG, NER, and LLMs for Instagram influencer identification
• Consulting: Delivered ML solutions for search engines, DeFi agents, and predictive markets for blockchain clients (Solana, Polygon)
• Innovation: Built proprietary stack with distributed crawling and network analysis for scalable real-time insights

• Led ML R&D team of 5 engineers, delivering systems across 1,150 sites serving ~200M monthly unique users
• Created the company's first distributed ML search engine, achieving ~16% traffic increase and ~25% CTR on over 1 billion threads
• Built novel forum representation models using Transformers & Graph Neural Networks for cross-domain content
• Drove ~18% YoY traffic engagement growth, supporting the company's successful IPO
• Engineered: model benchmarking & monitoring for A/B/n testing, live model swapping for Vespa.ai, and ELT pipelines

Areas of Applied Research:

• Entity linking, text summarization, graph & text representation learning
• Recommender systems, search (LTR/IR/Deep), social network analysis
• Anomaly detection & spam filtering

• Improved search quality with re-engineered SOTA architectures, reaching 5-13% increase in MAP
• Built proprietary ranking models (Q&A MLMM, DeepRank-inspired), increasing critical legal case retrieval by 22% (R@50)
• Deployed deep learning-based extractive question-answering models in production
• Led full-stack development with Terraform, Docker, AWS, Solr, Elasticsearch, Python