Hi, my name is

Sourav Karmakar.

I build algorithmic trading systems.

I am a Quantitative Developer and Algo-Trading Engineer specializing in low-latency execution engines, options automation, and ML-driven market signals. Currently designing scalable financial infrastructure at Levitas.

Check out my code

About Me

Hello! I'm Sourav, a developer who bridges the gap between software engineering and financial markets. I enjoy engineering robust systems that can handle the complexities of market microstructure and real-time data.

My expertise lies in building Backtesting Frameworks, Execution Engines, and predictive models using Python and C++ logic. I have successfully deployed strategies achieving over 93% prediction accuracy in controlled environments.

Here are a few technologies I've been working with recently:

Python (Pandas, NumPy)
DuckDB & PyArrow
Pine Script
Machine Learning (LSTM/CNN)
Market Microstructure
Vectorized Backtesting

Where I've Worked

Software Developer (Quant & Trading Systems) @ Levitas

Feb 2025 - Present

Built a multi-client, multi-stock Cash Execution Engine optimized for low-latency routing and dynamic position management.
Engineered a production-grade Backtesting Framework using Python, Pandas, PyArrow, and DuckDB for large-scale historical simulations.
Developed an Options Execution Engine with fast option-chain parsing, strike-filter logic, and automated SL/target adjustments.
Implemented ML-based prediction modules (LSTM, ensemble models) for trade decision support.

Technical Projects

Backtesting Engine

A modular engine supporting OHLCV ingestion, bar-level event loops, vectorized PnL calculation, transaction costs, and slippage modeling.

Python Pandas Vectorization

Options Automation System

Implemented automated entry/exit logic, chain filtering, IV-based decision making, and real-time multistrike execution monitoring.

Algo Trading Options Automation

Market Data Pipeline

Designed fast CSV/Parquet ingestion using PyArrow + DuckDB, enabling sub-second analytics on multi-million row datasets.

DuckDB PyArrow Big Data

Stock Prediction Models

Built ANN/CNN/LSTM pipelines achieving 93%+ accuracy in hackathons; implemented feature engineering and drift handling.

TensorFlow LSTM Deep Learning

Bike Sharing Demand

Predicted demand with R² 0.909 using Gradient Boost. Featured in a published book.

ML XGBoost Research

Airline Referral Forecasting

Achieved 95.7% accuracy with XGBoost, SVM, and RF; performed segmentation and feature importance analysis.

Classification Analytics Sklearn

Publications & Achievements

Research Publications

Karmakar et al. (2025). Big Data-Driven Smart Climate Change Prediction Using ML. IJIRT.
Karmakar et al. (2024). Predicting Bike Sharing Demand. Routledge.
Karmakar et al. (2024). Book: Neurolink, Lambert Publication.

Honors

93%+ Accuracy Stock Prediction (Hack IT Sapiens Hackathon)
3rd Place CU Project Expo for DL-based forecasting
Global SAS Certification