Technical Projects

---

Bioinformatics

DNA Sonification Tool
Fasioned an auditory display tool in R for genetic sequence analysis by designing algorithms incorporating music theory to convert nucleotide and protein sequences into songs. Enhanced the musicality of the tool’s output while maintaining the overall analytical capabilities. (October 2020)

Code

Genetic Engineering Attribution
Developed algorithms to identify the most likely lab-of-origin for genetically engineered plasmid samples using machine learning techniques and features extracted from the nucleotide sequences. The developed algorithm outperformed the existing state-of-the-art model in use. (September 2020)

---

Healthcare Analytics

DengAI - Disease Spread Prediction
With the aid of environmental data collected over a decade in two South American cities, ensemble methods were developed to forecast the spread of dengue, using time series analysis and statistical modeling. Trained the models on Collab GPUs and acquired a top 5% rank globally. (July 2020)

Code

Flu Shot Learning
Employed deep learning techniques to predict whether people got H1N1 and seasonal flu vaccines, using sentiment and behavior data from the National 2009 H1N1 Flu Survey conducted by the CDC. mplemented a Neural Factorization Machine to attain an AUC of 0.86 and got a top 10% rank. (June 2020)

Code

---

Miscellaneous

Climate data analysis - Department of Chemical Engineering, IITM
Analyzed the data collected by moving and stationery sensors placed across multiple cities in India. Performed extensive data cleaning and validated key insights through hypothesis testing. Created temporal visualizations and geospatial heat maps for radiation and air quality-related parameters. (August 2019)

Traffic Management - Sangam ML Hackathon, 2019
Developed statistical models to aid in traffic management. Built a deep neural network to predict traffic volume using traffic and climate data collected over four years. Achieved an accuracy of 93.9% and finished as one of the top 10 among the 450 participating teams. (July 2019)

Fraudulent Transaction Prediction – Department of Computer Science, IITM
Designed a random forest classifier to predict fraudulent transactions. Conceptualized an online fraud prediction algorithm and won the first position in the Exebit Data Science Challenge. (April 2018)

Code

User Preference Modelling
Ideated and created a website to work with dynamic test data. Created logistic regression models to predict probabilities of the user’s phobias using correlating parameters from a survey conducted in the UK. (February 2018)

Code