-
π I'm currently working on:
- Sketch2Graphviz, a CLIP + Llama VLM fine-tuned with LoRA to convert sketches of graphs to Graphviz code
- Interpreting and Steering Over-refusal in LLMs with Refusal Tokens, π Accepted to the 2025 NeurIPS workshop on Mechanistic Interpretability π
-
π Winner of the 2023 Congressional App Challenge with TruthGuard, an AI-powered fake news detector
-
π― I'm looking to collaborate on research projects/papers and impactful AI-powered projects.
- Feel free to reach out and discuss any ideas: rishabsaia@gmail.com
-
π¨βπ» All of my projects are available at https://rishabalagharu.com/
-
π Resume
- Atlanta, Georgia
-
08:55
(UTC -05:00) - https://rishabalagharu.com/
- in/rishab-alagharu
- https://medium.com/@rishabalagharu
- https://huggingface.co/RishabA
- https://leetcode.com/u/RishabSA/
Pinned Loading
-
interp-refusal-tokens
interp-refusal-tokens PublicWe study whether categorical refusal tokens enable controllable and interpretable safety behavior in language models.
-
Sketch2Graphviz
Sketch2Graphviz PublicConvert sketches to Graphviz code with a Vision Language Model (CLIP + Llama 3.1) fine-tuned with LoRA.
Python 6
-
TruthGuard
TruthGuard PublicTruthGuard is an AI-powered fake news detector that detects misinformation and bias to combat the spread of misinformation. TruthGuard empowers you to separate fact from fiction and protect the truβ¦
Jupyter Notebook 6
-
PoliticalPulse
PoliticalPulse PublicPoliticalPulse is an AI-powered tool that gives members of Congress, their staff, and constituents a real-time read of public sentiment. By scanning trusted sources, it summarizes public opinions aβ¦
Jupyter Notebook 6
-
-
advanced-deep-learning
advanced-deep-learning PublicThis repo contains the notebooks and projects that I made to learn about different deep learning concepts in PyTorch
Jupyter Notebook 6
If the problem persists, check the GitHub status page or contact support.