I'm a PhD candidate in Physics at MIT, where I'm part of the Binary Star Astrophysics group at the MIT Kavli Institute. I work on machine learning for time-domain astronomy — specifically, self-supervised representation learning on irregularly-sampled light curves from large surveys like the Zwicky Transient Facility (ZTF).

📣 News (2026): Template Matching, Not Time Learning accepted at the AI4Physics workshop @ ICML 2026 — code & paper »

My research sits at the intersection of:

• 🌌 Ultracompact white dwarf binaries — extreme stellar systems with sub-hour orbital periods, important as LISA gravitational-wave sources and Type Ia supernova progenitors
• 🤖 Representation learning — transformers, contrastive learning, and interpretability of learned embeddings on real-world scientific time-series
• ⚡ Scalable ML systems — distributed multi-GPU training, large-scale data pipelines (TB-scale), and GPU-accelerated periodicity search

• period-diagnostic — Template Matching, Not Time Learning (accepted at AI4Physics @ ICML 2026). A drop-in diagnostic that decomposes period-regression R² into between- vs. within-class signal, revealing that self-supervised light-curve encoders recover stellar periods by class-template matching, not genuine time learning — across eight encoders from a 4.4M BiGRU to the 110M MOMENT foundation model. 📄 paper
• ztf-embedding-demo — interactive demo for exploring learned ZTF light-curve embeddings; clickable sources, comparison across representation spaces, focus on whether embeddings capture physical structure vs. survey systematics
• Self-supervised transformer for ZTF — dual-head architecture trained with contrastive learning on 1.4B+ irregularly-sampled light curves; scaling to multi-GPU DDP on 16 A100 GPUs on MIT's Engaging cluster (MGHPCC, Holyoke).

• Chickles, E. & Audenaert, J. "Template Matching, Not Time Learning: A Diagnostic for Self-Supervised Light-Curve Encoders" — AI4Physics Workshop, ICML (2026) · code & paper
• Chickles, E., et al. "An eclipsing 8.56-minute orbital period mass-transferring binary" — ApJ (2026)
• Chickles, E., et al. "A gravitational-wave–detectable Type Ia supernova progenitor" — ApJ (2025)

• 📧 echickle@mit.edu
• 🌐 emmachickles.github.io
• 🔭 ORCID