Experience

Machine Learning Intern

Contract (Confidential)

• Designed and trained a neural network using letter, phoneme, and metaphone sequences with Embedding-LSTM modules and an MLP, achieving 90.7% accuracy and 93.59% precision on string similarity classification.
• Built a dataset of 4000+ labeled string pairs and engineered feature extraction pipelines including tokenization, phoneme and metaphone generation, sentence embedding cosine similarity, and Levenshtein distance.
• Deployed an ONNX-optimized model into a Flask API for real-time inference, accelerating predictions by 4x through preprocess batching, fuzzy match pruning, JIT compilation, and C++-backed operations.

Undergraduate Researcher

UCF Center for Research in Computer Vision

At the UCF CRCV Lab, I worked with advanced computer vision models to improve 3D pose estimation tasks and apply them to quantify human movement control for patients.

• Built a custom dataset of 1,036 video samples of stroke patients performing Box and Block Tests, segmenting videos into 30-frame clips for temporal action classification.
• Fine-tuned and benchmarked neural networks (R3D, R2Plus1D, Video Swin Transformer, Video MViT, MotionBERT, PoseConv3D, MS-G3D) for movement analysis, achieving up to 90.18% accuracy across different seeds.

Software Engineering Intern

Dynamic Animation Systems

My first software engineering internship was at DAS, where I worked with a range of technologies, from Hugging Face and Langchain for Natural Language Processing and LLM training, to JetBrains MPS, the Truffle Language Implementation Framework, and ANTLR for designing and building domain-specific languages.

• Fine-tuned the Mistral-7B LLM with Hugging Face's Transformers and PEFT libraries to generate simulation scenario files compliant with an XSD schema.
• Developed a graph-based ordering system to manage transactional processes in a declarative rule-based engine.
• Designed an ontology for simulation hosting, enabling deployment in on-premises and cloud environments using Docker and Kubernetes, with support for AWS and GCP.