About the author
Software engineer focused on useful, explainable AI systems.
I’m Sai Prasanna Maharana, a software engineer in Bengaluru, Karnataka, Indiawith 7 years of professional experience. I use this hub to turn ongoing AI study into connected explanations that can be checked, applied, and revisited.
Why this exists
Learning in public, without skipping the hard parts.
My work and study sit where retrieval, agent orchestration, evaluation, and cloud delivery meet. The useful part is not merely making a demo respond—it is understanding why a system behaves as it does and how to test whether it is dependable.
Notes hold durable explanations. Discovery captures dated observations and changes. Keeping those two forms separate makes the learning trail honest while still connecting new findings to stable foundations.
Current learning focus
The questions I’m working through now.
- 01Agentic AI
- 02Retrieval-augmented generation
- 03LLM evaluation
- 04Reliable cloud delivery
Explore by topic
Expertise grounded in actual notes.
Selected writing
A small sample of the work.
- 01NoteAttention mechanism in detailA step-by-step account of queries, keys, values, masking, scaling, and weighted sums.
- 02NoteMulti-head attentionHow parallel learned projections create complementary attention views.
- 03DiscoveryNumeric formats in TransformersA concrete comparison of FP64, FP32, and quantized INT8 representations.
Connect
Find the public work, or say hello.
These are the public profiles and contact address supplied for this project. No contact form or account is required.