Predictive Modeling & Research

[Python] [TensorFlow/Keras] [Computer Vision] [Statistical Analysis]

This project moves beyond standard library implementations to demonstrate proficiency in designing, training, and analyzing custom Convolutional Neural Networks (CNNs). It focuses on the mathematical impact of architecture choices on classification accuracy and training stability.

• Architecture: Designed custom CNNs using the Keras Functional API with Batch Normalization (BN) to manage internal covariate shift.

• Optimization: Benchmarked Adam vs. SGD optimizers to determine the most effective strategy for convergence on low-resolution data.

• Application: Selected this specific architecture for an Electric Vehicle detection system due to its superior spatial hierarchy handling.

Ready to discuss specific advantages regarding feature invariance and computational efficiency.

Enterprise AI Agents

[LLMs] [LangChain] [Jira API] [Automation]

I bridge the gap between Strategic Analytics and AI Automation. Beyond traditional predictive modeling, I design and prototype Agentic AI systems—intelligent workflows capable of reasoning, tool use, and task execution.

What I Build:

• Smart Orchestration: Systems that can translate unstructured business emails into structured Jira tickets or technical tasks.

• Legacy Code Refactoring: Workflows that ingest legacy scripts (SAS/R), map the logic, and draft modern Python replacements.

• Governed Digital Workers: Automated agents that perform low-level data operations (QA checks, report generation) under strict human supervision.

I combine the statistical rigor of a traditional Data Scientist with the engineering mindset of a developer to build AI that works safely in the enterprise.