Rugved Chandekar

A technical blog post describing four production engineering techniques (compact prompt format TOON, batching to remove LLM-from-loops, RAG for selective memory, and result caching) used to reduce LLM token usage from ~200k to ~5k per pipeline iteration and improve runtime.

Research paper presenting an uncertainty-aware attention-augmented DenseNet121 for pneumonia detection from chest X-rays, integrating Monte Carlo Dropout for uncertainty and Grad-CAM for explainability; listed as an IEEE conference contribution (2026).

An essay on engineering philosophy emphasizing deep understanding ('why' over 'how') and how that approach produced practical optimizations like token reduction and throughput improvements in production AI systems.

A write-up describing the design and usage of the integration-smoke-test PyPI package for diagnosing API connectivity issues across DNS, TCP, TLS, and HTTP layers.