Projects
- Music Collaborative Filtering
- This repo features the code for obtaining artist recommendations based on user’s history of artist likes.
- Matrix Factorization using Alternating Least Squares is used here.
- {Code}
- YouTube AI Assistant
- Ask questions about any YouTube video to this LLM powered assistant.
- This web-app uses LangChain and Streamlit.
- {Code}
- Pets Name Generator
- With the help of OpenAI’s apis, this web-app generates pet names for your pet based on the animal type and its skin color. Pet animal and pet’s color is provided as input by user.
- Using LangChain’s LLMs, Prompts, Chains and Agents
- {Code}
- Adversarial QA (powered by SQuAD)
- Built an interface that allows end-users to craft adversarial paragraphs, given a SQuAD (dev-set) paragraph, and test the adversarial on the language model BERT, which runs in the back-end.
- Successful adversarials (where model predicts incorrectly) can be submitted to the system via the button ‘Submit’.
- {Interface} {Code}
- Brain Networks - Comparison & Analysis
- Performed structural analysis on human connectome and chimpanzee connectome data.
- Implemented graphical measures such as centralities, motifs, clustering coefficients, max-flow, analysis of unique edges, etc. among others, and showcased differences between the two brain networks.
- {Slides} {Code}
- Pattern Studio using Deep GANs
- Built a system that generates floral and geometric design patterns using self-attention GANs, and applies the generated pattern on the garment selected in the image of a user.
- Implemented in Python (PyTorch).
- {Slides} {Report}
- Music Generation for given Poetry Lyrics
- Developed a neural model using stacked LSTM layers which takes poetry text as input and generates a suitable melody for it as the output.
- Implemented in Python.
- {Slides} {Report}
- Text-Ninja
- An application that can be used to find similar paragraphs across single or multiple text documents.
- An Android app.
- {Code}
- MyMemo - TimeTable managing Personal Assistant
- An Android app takes as input your activities throughout the week, for e.g. your classes, your TA hours, badminton time, etc.
- Based on the input information provided, it finds the vacant time available for work in the week and generates two options of timetables, which have specific times for ONLY work (research).
- It also reminds you of the next task (whether work/others) for the day.
- {Download App} {Code}
- Unibrowser
- An optimized search engine only for universities and college towns.
- Deployed as a web-app.
- {Code}