|
Vaibhavi Singh I am a graduate student in Computer Science at NYU Courant, specializing in problem-solving and reasoning in language models. I study how language models solve complex problems through step-by-step reasoning, focusing on mathematical and logical tasks. Prior to NYU, I engineered large-scale software systems at Adobe and Salesforce, building the core libraries that power Creative Cloud and Einstein AI for millions of users. Most recently, I developed production NLP systems and clinical risk prediction models in healthcare. |
Research |
Understanding Reasoning Failures to Improve Problem-Solving in VLMs
Chain-of-Thought Reasoning, Evaluation Frameworks, Vision-Language Models
Developed a systematic evaluation framework to identify why vision-language models fail at multi-step reasoning tasks. Created a failure taxonomy distinguishing perception errors from logical decomposition failures, revealing architectural bottlenecks that limit problem-solving capabilities. Paper Code Slides |
Information Density vs. Model Scale: Tradeoffs for Downstream Reasoning
Scaling Laws, Data-Efficient Learning, Foundation Models, Tokenization
Investigated architectural bottlenecks in representation learning that limit performance on complex reasoning tasks. Through systematic ablation, identified tokenization granularity as the primary constraint—4× finer encoding (144 vs 36 tokens/image) outweighed model depth and dataset scale for fine-grained problem-solving. Paper Slides |
Academic Service |
|
NeurIPS 2025
Ethics Reviewer — Datasets & Benchmarks Track
|
Experience |
|
Machine Learning Engineer
Healthcare AI Startup, India 2024 – 2025 Built clinical risk prediction models (XGBoost, TCN) achieving 0.87 F1-score through feature engineering, SMOTE for class imbalance, & hyperparameter optimization. Processed sparse EMR data for early-stage healthcare applications. |
|
Software Engineer II
Salesforce, India 2023 – 2024 Engineered petabyte-scale data ingestion pipelines, reducing latency by 30% for Einstein AI & real-time analytics. Scaled multi-tenant Kubernetes infrastructure on AWS for 200+ microservices. |
|
ML Systems Engineer (MTS II)
Adobe, India 2021 – 2023 Optimized heterogeneous compute (CPU/GPU) architectures for on-device neural inference, reducing latency for 20M+ users. Extended core C++ text-processing engines to handle complex document analysis and font parsing, ensuring high-throughput performance under strict SLAs. |
|
Cloud Infrastructure Engineer (MTS I)
Adobe, India 2019 – 2021 Scaled distributed data serving infrastructure, optimizing high-throughput request handling for 10M+ daily users. Reduced compute overhead by 12% through system-level performance profiling. |
Education |
|
M.S. in Computer Science (Machine Learning)
New York University, Courant Institute 2025 – 2027 (expected) GPA: 3.89/4.00 Research focus: problem-solving and reasoning in language models Coursework: Natural Language Processing (Eunsol Choi), Deep Learning (Yann LeCun), Computer Vision (Saining Xie) |
|
B.E. Computer Engineering (Hons)
Netaji Subhas Institute of Technology, University of Delhi 2015 – 2019 First Class with Distinction
Graduated in the top 10% of the department
|
|
Last updated: February 6, 2026
|