Developed a large language model for Graph Protocol Uniswap v3, implementing optimized GraphQL queries and the llama-index library to efficiently manage extensive texts and facilitate integration with OpenAI's API.

Conducted governance research with uniswap and a16z through collegeDAO

Conducted on chain analysis and developed 10 Dune or flipside dashboards of shock events spanning across DAOs in the NFT and DeFi space like Nouns, Uniswap, Pudgy penguins, Farcaster and more.

Won ETHOxford hackathon bounty by developing an L2 using Polygon zkEVM and fork uniswap v2 and ran transactions through the DEX to stress test the L2.

Conducted textural analysis on proposals spanning across shock events in Decentralised Autonomous organizations (DAO) across the Non Fungible Token and decentralized finance space such as Nouns and Uniswap

Identified metrics accessing significance of proposals affecting acceptance in DAOs as input for Large Language Model

Built a large language model prototype for graph's uniswap V3 to get students to learn how to build subgraphs and substreams.

Leading a research project that analysis web2 to web3 brands growth, starting with scraping on chain and off chain data from niftygateway to conduct analysis on their growth

Deep Reinforcement learning model for stock trading

Used OpenAPI gym to build a long-term investing model that outperformed S&P 500 

Causal Impact of Healthcare Intervention on Spending and Coverage 

Conducted a rigorous causal inference analysis using regression models in R to evaluate the impact of healthcare interventions on spending and coverage, successfully distinguished between treatment and control groups and quantified policy effectiveness. 

Project Caterpillar: Developed models for freight and inventory optimization involving complex logistics.

Worked on data consolidation from 150+ ERP systems, reducing costs and improving efficiency in parts procurement and distribution, learning Genetic Algorithm using Python (Pandas) and ETL techniques for data manipulation and preprocessing.

Project Accident Occurrence predictions with CSFS (Cost Effective Fisher Score)

Extract real time traffic data for data cleaning and manipulation with Python. Used feature selection methods like Cost-sensitive SVM Classifier and my own CSFS to find key variables to solve the problem of cost imbalance and noisy data feature selection. 

SEC Filing Similarity Indexing for Trading Signals

Coded in Wolfram Mathematica to analyze and compare the language and word usage in SEC filings from consecutive years for large-cap technology companies. Investigated the correlation between the similarity of language used in these filings and the companies' stock performance, independent of market returns.

Traffic accidents are a huge social cost that needs to be curbed and prediction of traffic accidents helps officials in implementing strategies for safer roads. Most of the research in this field focuses on using powerful classifiers for higher performance in making traffic accident predictions. This report works on building a high performing model by introducing cost-sensitive feature selection methods as a pre processing step that helps to curb the class imbalance and noisy features problem most accident prediction models face when only a classifier without a pre-processing step is used. 

This study introduces three cost-sensitive feature selection methods that factor in imbalanced data and helps in figuring out important features using various performance metrics. The selected features are then used in the Cost-sensitive Support Vector Machine Classifier to check for the performance. K-Nearest Neighbor classifier was used as well to draw comparisons on the performance of the three feature selection methods. Overall, the report has found feature selection to be an effective pre-processing step for traffic accident prediction and of all features used, weather was found to be the most important feature with the highest feature score and AUC score, followed by features such as location, time and speed. The results were comparable to previous researches in determining important features for traffic accident prediction.

Efficient freight and inventory optimization lie at the heart of modern supply chain management, offering organizations the promise of reduced costs and enhanced operational efficiency. This research paper delves into the intricacies of "Project Caterpillar," a transformative initiative focused on developing advanced models for freight and inventory optimization.

In the backdrop of complex logistics, this project undertook the monumental task of consolidating data from over 150 Enterprise Resource Planning (ERP) systems. The objective was clear: to streamline the procurement and distribution of parts, thereby optimizing resource allocation and driving cost reductions.

At the core of this endeavor is the implementation of Genetic Algorithm (GA), a cutting-edge optimization technique. Leveraging Python (Pandas) and Extract, Transform, Load (ETL) techniques for data manipulation and preprocessing, the project meticulously crafted an intricate framework. This framework was designed to adapt and thrive in the dynamic world of logistics, where real-time adjustments and intelligent decision-making are paramount.

More details on the results will be provided soon.