DeFi (Decentralised Finance) Bank Dapp - Ethereum, Smart contracts

• Decentralised (DeFi) Bank Dapp on traditional banking products
• Built robust Smart contracts using Solidity programming.
• Setup end to end project setup on blockchain using ganache.
• Token and defi bank Smart contracts deployment to Ethereum.
• Built DeFi Dapp allows users to connect & interact with the smart contract.
• Created unit tests using JavaScript to test smart contracts
• Code base available on GitHub and Dapp deployed to GitHub.

https://github.com/itznishant?tab=repositories

Skills: Solidity, DeFi, Dapps, Smart Contracts, Ethereum, Git, Hardhat, React JS, Web3.JS, Ganache, Metamask

• Built on robust smart contracts in Solidity
• Instant crypto swap Exchange with no time wait needed to fill orders.
• Test driven development (JS) using Mocha test framework / Chai.
• App Frontend built with React JS components and HTML.
• Displays ETH & Token Balance in app UI
• Displays account details in the UI along with custom avatar.
• Tests built using Mocha / Chai assertion libraries.
• Basic tests for token & token sale smart contracts deployment.

Election Voting DAPPusing Ethereum Smart contracts

• Election voting DAPP (Decentralised Application) built on Ethereum.
• Smart contract for the project built using Solidity programming
• DAPP allows accounts to vote for listed candidates in an election.
• Tests to validate for double voting & voting for invalid candidates.
• Built on Ethereum Blockchain using Smart contracts on Solidity.
• Front end web 3 app interface to vote for listed candidates & view election results.
• Smoke tests / Unit tests using truffle test framework/ JavaScript & smart contracts deployment with truffle migrate
• Integrating accounts using Metamask

Skills: Solidity, JavaScript, Smart Contracts, Ethereum, Git, Truffle framework, Ganache, Metamask, HTML, web 3

GSNA GTRF CMB Scoring:

Description:

GTRF is a trade finance product live in multiple countries. The global scoring model for GTRF tracks transactions and customer behavior based on typologies and scenarios to identify risky customers & counterparties with an objective to reduce suspicious cases as well as False Positives. This is extremely critical to business as this represents customers at highest risk.

Goals:

Reduce suspicious cases / False Positive of the scoring model by using multi-dimensional data of the customers.

EC (Entity Consolidation):

Finding over linked / under linked entities for customers in multiple countries.

Responsibilities:

• Completion of assigned sprint developmenttasks.
• Development / Enhancements of existing scenarios and typologies.
• Extensively used Python,Scala on Apachesparkonenterprise data lake.
• Collaborating with Domain SMEs to tune threshold by BTL/ATL Testing.
• Presenting demos to business stakeholders.
• Involved in IMR review process to prepare model documentation.

Description:

A retail firm wants to set up supermarkets in Hyderabad city but is not sure about the location. Chosen locations should ideally have a considerable population, near to work centers/residential districts for easier access to a large number of citizens. Business objective is to find suitable neighborhoods to open stores.

• Extracting house prices data for neighborhoods using web scraping libraries like BeautifulSoup and converting to data frame using Pandas.
• Getting geographical coordinates (latitude, longitude) for neighborhoods using geojson data and creating maps for visualizing using Folium.
• Filtering relevant neighborhoods using python Boolean filtering.
• Capturing Venue information for neighborhoods using Foursquare API.
• Converting categorical venue categories to numeric by one hot encoding.
• Finding optimal k using elbow method and clustering neighborhoods using most common venues data using k-means clustering.
• Visualizing the clusters on the Folium map and identifying patterns and suggesting suitable neighborhoods after examining the clusters.

Description:

Customer Churn Analytics for a UK based Retail firm.

Business problem: Predict the customers likely to churn based on historical transactional data and demographic data.

Responsibilities:

• Understanding of business problem to identify the outcome to be achieved.
• Extracting sample data from AWS S3 into Python using pandas.
• Data cleaning by renaming existing columns and dropping un-necessary columns.
• Analysis of numeric variables using pandas describe method.
• Identifying key columns (data attributes) using correlation plots with target variable.
• Creating derived features from data (ex: Frequency, Recency, No. of transactions etc.) to increase predictive power of the churn model.
• Identifying outliers using box plots and missing values (NaN) treatment using dropna, fillna functions, data transformations wherever required to enhance data quality.
• Model building using sklearn and evaluation using sklearn.metrics like precision/recall and log loss.
• Presenting findings to business through graphs and plots.

Responsibilities:

• Extracting text data from Documents into python.
• Building ML classification algorithm from scratch to classify documents in IC role.
• Training model for baseline acceptable evaluation metrics score.
• Collaborating with functional team for training data and labels.

ML Algorithm for Classification of documents:

Features:

• Multiclass classification models used.
• Train/test split to reduce overfitting of the model on training data.
• Python library used for data extraction from pdf documents to text.
• Data Randomisation with python random function.
• Wide array of Python data structures like lists, tuples and sets used.
• Made use of string functions, regular expressions, lambda (single line functions) for data cleaning.
• Lemmatization (root word) used for improving quality and reduce text features.
• Count Vectorizer and Vocabulary used for feature selection.
• Model trained to predict class with 92+ % score.
• Evaluation using Confusion matrix, classification report and accuracy score.
• Matplotlib and seaborn libraries used for visualizing data as graphs/charts and in confusion matrix.
• Predicted probabilities for each document label along with label inoutput.
• Model deployed as web service API and consumed from application.