NLP thesis focused on learning intermediate representations for question answering. Uses word vectors, text to text transformers, coreference resolution and other NLP techniques to generate inferred knowledge graphs from a passage of text.

Conference presentation for MLOps Community about our work to build clinically validated models to identify and treat suicidal ideation in healthcare settings.

Fine-tuned LLM agent that can perform RAG based chat completely in browser. Runs completely in a user's browser or chrome extension and can use tools with a ~1 billion parameter model.

Built a system to identify imporant passively sensed features for remote patient monitoring, including insights & interventions. Filed patent with core intellectual property.

Leveraging object recognition models to analyze traffic patterns from NYC's free traffic camera API.

Created a model to predict rent prices and apartment occupancy from relevant features. Data will be used to competetively price apartment rent and predict occupancy for profit analysis. Resulted in a model with less than 5% error rate using ensemble machine learning methods.

Programming Language created in less than a month for the REPL.it coding Jam. Features type inference, functional evaluation and C-like syntax

Python library that makes it easy to annotate and track physics objects in videos. Simply supply a configuration file and video, and the model will output annotated footage that can track objects and their velocities along with metadata accounting for camera tilt/pan.

E-Sports Statistics Platform focused on improving player performance through informed statistics. AI/ML Video Recognition technology used to analyze performance.

Context aware chatbot trained to answer questions from Wikipedia articles

Created tool to automatically clean and visualize data, and build machine learning models.

Language Models have been used for a variety of tasks- text generation, translation, etc. They have shown incredible strengths as few or zero shot learns. Let's take advantage of that to build a Chess Engine using GPT-3.

NodeJS library that can simulate quantum algorithms on traditional hardware. Functional DSL includes all common quantum gates (Hadamard, CNot, etc) and is highly extensible. Can be utilized to prototype quantum algorithms before exporting to costly hardware.

Chatbot framework for construction of intelligent agents. Can be used to offload service requests with easy to use syntax.

Drawing app that utilizes machine learning and handwriting recognition to evaluate mathematical expressions. Also supports graphing and mathematical variable assigment and numerical computation.

Codely is an integrated web IDE that allows for the live coding of JS apps. Code can be shared by simply copy/pasting the URL on any codely page.

Smart chart rendering for large sets of data. Preserves maxes, mins and other general trends on zoom.

Auto Website
Kit builder for after-market transmission controllers and various car modificationsa.

An online IDE that allows easy editing of git repositories stored on Github.

Genetic algorithm library for native JS. Easily abstract complex genetic algorithms into simple wrappers for machine learning and AI projects.

A mind-bending game built around a physics engine resembling the famous M.C. Escher's artwork.