Research Experience

Center for Visual Computing, UCSD

Advised by: Prof. Manmohan Chandraker

• Project: Foundational Model for Texture, Illumination and Spatial Reasoning (Ongoing)
  • Vision Language Model to infer scene level object and lightning characteristics

Center for Applied Internet Data Analysis, UCSD

Advised by: Prof. KC Claffy, Ricky Mok

• Project: Botnet Scan Event Detection (Ongoing)
  • Benchmark to analyze internet-wide vulnerable port scanning activities that are precursors to DDoS attacks

MixLab, UCSD

Advised by: Prof. Zhiting Hu

• Project: World Model Benchmark
  • Contribution to the benchmark dataset that evaluates VLMs’ perception capability in inferring environment state change and cause of change
• Project: Efficient Concept Learning using Causal Inference
  • Contribution to the development of a few-shot learning framework, based on causal discovery, to categorize objects through visual concepts

Department of Chemical Engineering, BITS Pilani

Advised by: Prof. Suresh Gupta

• Project: Fast Gas Concentration Prediction in Dynamic Environment
  • RNN-based algorithm to predict concentrations in a tertiary gas mixture within a dynamically changing environment using TGS sensor array

Advised by: Prof. Krishna Etika

• Project: Optimization of Solar Energy Absorption by NanoFluids using Genetic Algorithm
  • Computational model based on genetic algorithm to determine the optimal nanofluid concentration and flow rate that maximizes energy absorption at a given solar incidence

---

Open Source Projects

DiTP: Transformer-Based Diffusion Model for Trajectory Planning

Long-horizon path-planning algorithm using diffusion-transformers with SOTA performance on Maze2D environments

LLM-Embed: Increasing Performance of RAG using LLMs as Encoder

Zero-shot retrieval system using causal-LLMs and FAISS with 5x improvement in document retrieval performance