AI Product ManagerSashidhar Yammanur
Most people ship the magic and run blind.
I build the machinery underneath. Evals that gate before you ship. Traces that tell you why a message happened. Guardrails that hold under pressure. Cost attribution that ties to actual user value.
The feature is the easy part.
What I Build
There is a gap between the demo and the real product. Most teams find it the hard way.
I close it in four ways.
Evals
I build tests that catch problems before users do. Change a prompt, know in minutes if something broke.
Observability
I build systems that tell you what went wrong and why, without waiting for a complaint. Every decision, traceable.
Architecture
I design the full pipeline. What the model sees, how it retrieves, when it stops, what happens when it fails.
Cost and Safety
I track what the AI actually costs, down to the user journey and tenant. I build the guardrails before you need them.
Writing
How to read evals. How to design rubrics. Where pipelines break.
Kappa 0.78. The rubric was broken.
The headline number looked fine. Here's how to find what it's hiding.
Rubric DiagnosisLLM EvaluationTwo numbers. One question.
When human-human agreement and human-LLM agreement diverge, the gap tells you exactly what to fix.
LLM JudgeEval ArchitectureNever split the patient.
A common architecture decision that looks efficient and quietly makes your health AI blind to the findings that matter most.
Pipeline ArchitectureMulti-Signal AIYou can't evaluate what you haven't defined.
Most teams start building the eval before they've written a north star sentence. Here's the sequence that actually works.
Task EfficacyEvaluation Foundations
About
Enterprise software at Amdocs, shipping digital transformation for telecom operators across Europe and Asia. Rural India at AgroStar, building product for farmers at scale. Health and fitness at Fittr, where I handled all things product, revenue & marketing.
Today I run a consulting practice working with founders on AI product and growth, and Arkome Labs, a deep-tech multi-omics startup at the intersection of organ on chip, metabolomics and sequencing.