sachin bhoi

Data Analyst

Exponentia.ai•  February 2020 - Present

• I am a part of a four people team working on an application to automate the invoice generation process. I lead the Python part of the project. I designed a framework in python to integrate SQL database in python, configuration files. I made a nested JSON files by pulling data from configuration files and SQL database, validate the JSON files, and hit the API to generate a response, and store the response in the database. I have successfully integrated the modules in the application. I made the GUI in Tkinter for of the application. • I have worked on natural language processing for building chatbots, speech emotion recognition of voice, sentiment analysis, speech diarisation to distinguish audio of different persons.

Data science intern

Essar Oil•  October 2019 - January 2020

• Essar has been using Monte Carlo Simulations to forecast the margins of the refinery. I performed the Augmented Dickey-Fuller test and discovered that their data is mean-reverting. Monte Carlo Simulation is based on browning motion, which does not make the model ideal for forecasting a mean-reverting series. • Models with previous memory are the best fit for such data. I employed the “Seasonal Auto-Regressive Integrated Moving Average” (SARIMA) model, which is a statistical model. The model predicted 5 years in the future as a high and low with a 90% probability that the real margin will fall within the given range. The model was better at predicting extreme ends making their traditional approach obsolete. I also developed CNN to do the same.

CFD simulations

Airbus•  May 2019 - August 2019

• AIRBUS is conducting a fuel tank leakage test using a compressed air-helium mixture within the wing fuel tank of the Aircraft. It is a state of the art project with no previous research. The key requirement is to understand the dispersion and mixing of the helium-air mixture within the fuel tank during the leakage test and find a better strategy for quick, and homogeneously mixing the gases within the tank. • Manually generated a fine mesh up to 41 million cells with internal boundary layers. The skewness of the cells was extensively minimized. RANS k-w SST model was used because it solves the boundary layer and multiphase mixture model was used for the mixture of the gases. Performed multiple steady-state and transient simulations on 64 cores HPC. Grid Convergence Index study was carried out to validate the simulations. The results were computationally heavy and required 188GB of RAM and 32 CPU cores for post-processed on Tecplot360ex. • The CFD model forms a baseline for AIRBUS for prototyping and testing different configurations and parameters of the helium leakage test. In turn, I received distinction for the project.

Craanfield University

Computational Fluid Dynamics, MS•  September 2018 - August 2019

I did my masters in Computational Fluid Dynamics from Cranfield University, UK. My Individual research project was with Airbus on "Dispersion analysis of the helium-air mixture within the wing fuel tank of airbus a320". I graduated in 2019 with 4.0 GPA