Computer Science Graduate focused on Distributed Systems and Applied AI Enthusiast

Built and deployed academic-scale systems including Kafka–Spark streaming pipelines handling 5K+ events/sec and neural surrogate models achieving 92.4% accuracy with 90× simulation speedup.

Strong interest in backend engineering, data systems, and real-time analytics.

CS Major · Published at INCOFT-2025 · Open to SDE / ML / Data Engineering roles

Stream Processing · ETL Pipelines · Real-Time Analytics

Deep Learning · Model Optimization

RESTful APIs · Flask · MySQL · MongoDB · Transaction Management · Query Optimization

CI/CD

End-to-end streaming fraud detection system built using Kafka and Spark Structured Streaming.

• 5,000+ events/sec throughput · sub-2s end-to-end latency · 89% detection accuracy
• Hybrid rule-based heuristics and graph-structured anomaly detection using account relationship modeling
• Bronze → Silver → Gold layered data architecture
• Implemented indexed MySQL storage and optimized queries for low-latency analytics
• Real-time anomaly monitoring dashboard (PyQt5)

Technologies: Kafka • Spark • HDFS • Python • Graph Analytics

Neural surrogate model approximating CFD simulations of left-ventricle hemodynamics.

• 92.4% prediction accuracy · 90× computational speedup (45s → 0.5s)
• Predicts pressure, velocity, and wall shear stress fields
• Physics-informed custom loss functions preserving pressure–velocity consistency
• Modular inference pipeline with interactive Streamlit dashboard

Technologies: TensorFlow • Streamlit • Python • Surrogate Modeling

Computational fluid dynamics modeling of cardiac flow using ANSYS Fluent.

• Transient flow simulation with physiological boundary conditions
• Pressure, velocity, and wall shear stress contour visualization
• Hemodynamic behavior analysis under realistic cardiac cycles

Technologies: ANSYS Fluent • CFD • Fluid Mechanics

Super Resolution of Images Using Residual Networks
3rd International Conference on Futuristic Technologies (INCOFT-2025) · Feb 2025

Deep residual CNN for single-image super resolution — reconstructing high-frequency textures from low-resolution inputs using skip connections and perceptual loss, outperforming bicubic interpolation baselines on PSNR and SSIM metrics.

LinkedIn: https://linkedin.com/in/sinchana-s06
Email: sinchana.sridhar06@gmail.com
GitHub: https://github.com/sinchana-s-06