Manasi Sharma's Projects
robot
A curated list of Large Language Model (LLM) Interpretability resources.
CALVIN - A benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Tasks
Creates a web server that displays Star Trek images and and looks up queries in a table
These are the homework projects I worked on for the COMS W 4733: Computational Aspects of Robotics course at Columbia University (https://sites.google.com/view/coms4733-fall2020/home?authuser=0) in Fall '20.
Final Project for COMS W 4995: Deep Learning for Computer Vision course at Columbia University (https://www.deeplearningforcomputervision.com/home/deep-learning-for-computer-vision) in Fall '20. I created a film poster genre classification and captioning system (the ‘Genre-ator’) that used CNNs for image classification to identify the genre of a movie from its poster and used LSTMs to caption the poster with elements in the image (such as “a standing man in black”).
Final Project for the CS22W: Machine Learning with Graphs course at Stanford University (http://web.stanford.edu/class/cs224w/) in Autumn '21. We applied graph ML techniques and networks (such as the Graph Convolutional Network, GraphSAGE Network and Graph Isomorphism Network) to the ogbl-ddi dataset (https://ogb.stanford.edu/docs/linkprop/#ogbl-ddi) for drug-drug interactions.
Final Project for the CS238: Decision Making Under Uncertainty course at Stanford University (https://web.stanford.edu/class/aa228/cgi-bin/wp/) in Autumn '21. We optimized wind turbine placement in a wind farm, subject to wake effects, using Q-learning.
Code for the paper "Planning with Diffusion for Flexible Behavior Synthesis"
Codebase for the paper "To the Noise and Back: Diffusion for Shared Autonomy"
Predicting values of the cumulative number of confirmed cases, recovered cases and deaths for May 1, 2020. Part of Columbia's DSI's Challenge: https://datascience.columbia.edu/news/2020/dsis-center-for-health-analytics-launches-covid-19-data-challenge/.
An ML classifier for the Gattini-IR telescope (to classify sources as real/bogus), produced as part of the Summer Undergraduate Research Fellowship Program at Caltech in Summer '19.
prints hello world
Hierarchical Universal Language Conditioned Policies
Rapidly designing and solving differential games in Julia.
✴️ Avro schema for ZTF alerts