TestMu AI Blogs

Monika Sharma of Salesforce on DeepEval as pytest for LLMs, the RAG and safety metric taxonomy, choosing thresholds, and where eval tests fit in the pyramid.

A Testμ 2026 panel with CITY Furniture, Vail Resorts and TestMu AI on pooled QA teams, change failure rate, and why everyone becomes a QA architect.

Partha Sarathi Samal of Paramount on replaying 18 months of incidents against 4,200 tests, the proven, duplicate and unproven verdicts, and their limits.

Sobhitha Neelanath of Salesforce on why a 94.2% pass rate missed a $400,000 refund loop, and the trust equation quality leaders should measure instead.

Rhea Goel of Amazon on replacing a re-ranker with an LLM: natural language objectives, fine-tuning, DPO, hard and soft constraints, distillation and LLM judges.

Mesut Durukal of OMP on the ten root causes behind a four-hour pipeline nobody trusted, and the three AI skills his team now runs before any test merges.

John Liang of Salesforce on four phases of agentic testing, from validating every step to running thousands of concurrent voice calls with agent-to-agent tests.

A Testμ 2026 panel from Associated Bank, Symplyphy.ai and Fénix Advisory on why AI pilots stall on trust, probabilistic quality, and earning agent autonomy.

Deepak Kamboj of Microsoft on running 14,000 tests across a 100-package monorepo, risk-based selection that cut regression feedback 60%, and agent governance.

Amanda Martin of Vapi on three restaurant booking agents built from the same script, why only one caught an ambiguous date, and how to test with simulations.

Wessel Braakman of Bouvet on five developer types every tester meets, what each one needs from a bug report, and the encounters that taught him the difference.

Gil Zilberfeld of TestinGil on golden data sets, scorecards and sanity tests, and why fixing a bug inside a prompt is only the start of fixing it.

Sai Krishna and Srinivasan Sekar of TestMu AI on testing video AI agents: lip sync, false silence, turn taking, and one scenario across many personas.

Gurleen Kaur and Keerthivasan Devaraj of Piramal Finance on going from zero to 947 business builders, 369 live apps, and why quality became a hard CI gate.

Mahesh Sathiamoorthy of Bespoke Labs on why RL environments are the scarce ingredient in agent building, with case studies from Snowflake and Credit Karma.

Thomas Dohmke, ex-GitHub CEO, on loop engineering: agents writing and reviewing code, session logs as artifacts, token budgets and the next developer platform.

Four enterprise AI leaders on why generating 10x more code is the wrong goal, and what trust, determinism and outcome measurement actually take at scale.

Prince Verma of TestMu AI on why an agent sounds as confident when wrong as when right, and the six capabilities an agentic harness needs to ship safely.

Rahul Shetty on what QA engineers should learn first in the AI era, how to prove real agent experience to recruiters, and whether the SDET role is shrinking.

Nishant Gupta of Meta on the patterns that keep probabilistic AI agents safe in production: tool contracts, policy gates, budgets and replayable traces.