Hello! I Am Eric Zhou
A researcher who
Teaches machines
to 
understand us...
From speech to stories to reinforcement learning.
I'm a |
Currently, I'm an AI Engineer Intern atNew York Life
I'm an M.S. Computer Science student at Columbia University (B.S. from Georgia Tech) and a graduate AI researcher. I build retrieval-augmented and agentic systems, and study how language and speech models can be made more reliable — with past stints at Core42, Carnegie Mellon, and WW.
Work Experience
New York Life
May 2026 – Present
AI Engineer Intern
Building an agentic RAG system for automated information retrieval and insight generation, orchestrating workflows and evaluation on AWS Bedrock and Domino Data Labs, and benchmarking OpenSearch- and LlamaIndex-based retrieval pipelines across enterprise datasets.
Core42
May 2025 – Aug 2025
Software Engineer Intern
Engineered a 6-node highly-available Kubernetes cluster (3 control nodes, NGINX Ingress load balancing) supporting ~185 pods for GPT batch processing, fine-tuning, and inference, with full ELK logging and automated, reproducible deployments.
WW (Weight Watchers)
May 2023 – Aug 2023
Software Engineer Intern
Cut traffic to a core service handling 500k+ requests/min by ~12% in Scala, dropping a key endpoint's average response time from 22.1 to 18.4 ms and easing throughput pressure 14×, while writing unit/integration tests and mentoring new interns.
I'm currently looking to join a team building AI systems that matter
across machine learning, AI engineering, and quantitative research
Languages
Frameworks & Tools
Research & Publications
Speech · Pronunciation Assessment
OpenEnded: An Open-Response Speech Corpus for Pronunciation Assessment
A 1,000-utterance open-response L2-English speech corpus with expert annotations for accuracy, fluency, and prosody. Benchmarked eight LLM- and SSL-based scoring models (including Gemini audio-language models and wav2vec2/HuBERT), and fine-tuned a multi-task pronunciation model that raised mean correlation with human raters by 46%.
Chen, Y., Zhou, E., Ding, E., Shen, T., Zhou, Y. & Hirschberg, J. (2026).
NLP · Interactive Narrative
Generating Dynamic User-Guided Interactive Fiction Text Games
A game engine that converts stories into playable text games, using a neural planner and LLM prompting to translate live user input into dynamically generated game logic. The planner identifies causal relationships and parses stories into causal chains of events with connected preconditions and effects.
Zhou, E., Basavatia, S., Moontashir, S. & Riedl, M. (2025). Also: Chen, Z., Zhou, E., et al. — Ambient Adventures (WiNLP 2023).
Reinforcement Learning
Influence Functions for Offline Reinforcement Learning
A novel use of statistical influence functions to evaluate trajectories in offline RL 3–4 orders of magnitude faster than leave-one-out retraining. Validated on Gym environments with TD3/CQL, reaching 0.87 correlation with leave-one-out retraining while outperforming heuristic baselines such as return and TD-error.
