ABHIJEET SINGH

Real-Time Algo Trading Platform

• Built an end-to-end algorithmic trading system using Python to connect with broker APIs and fetch 1-second frequency stock tickers for all active NSE instruments.
• Designed system architecture on Azure Cloud, including ingestion, processing, execution, storage, and optional UI layers.
• Created a WebSocket-based Python client to subscribe to live feeds from multiple broker APIs (e.g., Zerodha Kite, Alpaca).
• Engineered a real-time signal generation engine to execute trading strategies like momentum, mean reversion, and breakout, using NumPy, Pandas, and TA-Lib.
• Built a modular order execution module with API-based order placement, authentication, and retry logic.
• Stored every-second ticker data and signal logs in Azure Blob Storage and Azure SQL for backtesting and auditing.
• Implemented Azure VM setup for execution, Azure Key Vault for secure API key storage, and Azure Monitor for real-time alerting and failure tracking.
• Designed and ran extensive unit and load tests, validating system performance under high-frequency data loads.

• Developed and implemented a data science solution for the mining industry using Azure ML, increasing productivity by 15% and reducing mining costs by $10 per metric ton.
• Migrated the data science infrastructure to Azure Cloud, enhancing productivity and efficiency through scalable cloud capabilities.
• Created predictive maintenance ML models for mining machinery, reducing machine downtime by 20% and saving $500,000 annually in operational costs.
• Built CI/CD pipelines for automating model training, registration, deployment, and creating endpoints for seamless integration.
• Extracted data from flat files and on-premise SQL servers to cloud storage, improving data handling and analysis, and identifying output changes, resulting in a $2 million annual revenue increase.
• Developed an automation system to categorize the incidents or service requests into different pre-defined categories using past dumps of data for healthcare diagnosis lab.
• Performed data pre-processing - stop word removal, lemmatization, emoticon & special character removal, contraction treatment.
• Trained LSTM based multi class classification model using only customer side conversation of historical data with 97% accuracy.
• Helped Service support team decrease user generated service tickets by 35%.
• Created Customer 360 Analytics Module for Insurance domain resulting in a decreased churn rate of 25% from 30%.
• Applied propensity model to generate 8% increase in high intent leads using XGBoost algorithm.
• Performed feature engineering to create features like CSAT score to get insights about customer satisfaction and support team effectiveness.
• Performed RFM analysis to better understand the customer and create customer segments.
• Used (logistic regression, random forest, decision tree, XGBoost) models for customer churn and evaluated their performances, used Boruta for modelling to identify important variables.

• Created and validated machine learning model for prioritizing which financial transactions should be manually investigated for potential money laundering.
• Performed EDA to understand the data and perform feature engineering to create new features.
• Used Unsupervised Learning (Isolation Forest) algorithm to create baseline models.
• Created classification model that could work in a cascading method on top of the Rule based system.
• Trained multiple classification models (Logistic Regression, Decision Tree, Random Forest, Knn) with different set of features and hyper-parameters.
• Created Random Forest classifier to capture suspicious transactions with a Recall of 96%.

• Delivered an end-to-end solution on WPD's Data Storage and BI reporting.
• Connected with business stakeholders to gather and analyse requirements.
• Designed and published dashboards in PowerBI with Standard Reports, Scorecards, drill down reports.
• Created UAT test cases; validate the quality and integrity of the data included in the Reports and Dashboards.