New York City Metropolitan Area
Managed a team building rule-based + statistical models for entity extraction (CRF, MaxEnt, SVM) Designed high-precision workflow to enrich financial text with entity metadata Oversaw internal tooling for feature engineering and large-scale model evaluation Collaborated with product teams to integrate extraction pipelines into real-time news systems
Developed CRF-based NER models using lexical, POS, orthographic features Built scalable feature extraction pipelines in C++/Python Improved precision/recall of entity tagging models through active-learning & model tuning Worked with linguists and analysts to refine domain-specific rules
Built text preprocessing components (tokenizer, POS tagger, sentence splitter) Implemented SVM/LogReg classifiers for topic tagging, document relevance & sentiment Optimized C++ pipelines for large-scale real-time text processing Collaborated with research teams to deploy statistical models into production
Supported applied machine learning coursework; assisted students with NLP assignments and deep learning fundamentals
• Fine-tuned semantic product matching model using human-labeled datasets (+10.2% recall) • Built multi-task relevance & similarity model (+2.8% MAP) • Filtered irrelevant products using transformer classifier (-17.2% noise)
• Identified risky shipping addresses via decision tree strategy (~290K monthly loss reduction) • Analyzed large-scale user/transaction data; unified inconsistent database structures
• Summarized theory for best subset selection and shrinkage methods • Simulated pure-noise datasets to evaluate inflation of significance levels
• Developed local alignment algorithm (EpiAlign) for epigenomic signal comparison • Implemented clustering models to classify 20,000+ tissue-associated genes • Built interactive web portal for EpiAlign visualization • Applied non-negative matrix factorization for mRNA isoform detection