Angle Up

projects.

I have worked on several interesting AI and data science projects. Most recently, I created a gemini-powered chatbot to aid with grant writing and other text-heavy tasks for Gallery263, a non-profit art Gallery in Cambridge, MA.

During my time at CMU I worked with Zachary Lipton on the speech-to-text tranformer model Abridge leverages to transcribe and capture noteworthy information from doctor-patient conversations. In addition to the python script linked below, an interim write up of this work can be found here.

ElevenLabs Voice Agent
Link icon
Voice Agent skeleton with tool calling (authentication, billing info), RAG policy-retrieval, and multi-agent orchestration.
STT TTS React.JS
Gemini-powered chatbot
Link icon
Streamlit UI & AWS EC2 MCP server leveraging hybrid search (BM25 + embeddings) context retrieval and gemini-powered tool-calling.
LLM RAG MCP
Clinical BERT
Link icon
Speech-to-text transformer model leveraging clinical BERT with context, in collaboration with Abridge.
BERT STT Fine-tuning
Sequence-to-sequence model
Link icon
End-to-end speech-to-text model using seq2seq architecture with attention.
RNN LSTM STT
Multi-layer perceptron
Link icon
Deep neural network (7 hidden layers + output) with Adam optimization and learning rate scheduling.
MLP Adam
RCT analysis
Link icon
Fixed effects and clustered SE regression analyses for an RCT with three conditions.
Regression Mediation RCT