Telemetry-Informed AI Agentic Testing (TIAT)

Yerram Sai Rakshit designed and led the enterprise deployment of Agentic Quality Engineering — a system in which autonomous AI agents generate, execute, prioritize, self-heal, and triage software tests while continuously reasoning over live production telemetry. Moving beyond brittle, script-based automation, the framework closed the long-standing gap between pre-release testing and post-release observability across 240+ microservices and 1,800+ test suites spanning roughly 50 releases per week. The result was a continuous, self-improving quality system that reduced test-maintenance effort by ~70%, improved defect-detection efficiency by ~45%, and cut mean time to detection by ~50% — outcomes at the leading edge of AI-driven quality engineering, consistent with the direction industry surveys such as the Capgemini–Sogeti World Quality Report identify as the future of testing. By unifying autonomous test generation, adaptive execution, self-healing maintenance, and telemetry-informed risk analysis into a single governed feedback loop, Rakshit operationalized a new discipline and advanced how reliable software is engineered at scale.

BUSINESS & TECHNICAL IMPACT
Engineering cost recovered: ~70% less test-maintenance effort, redirecting ~1,200 engineer-hours/quarter toward exploratory testing and quality strategy.
Reliability by design: Telemetry-informed coverage concentrates testing on the highest-risk, highest-traffic paths; self-healing resolves ~80% of change-induced breakages automatically.
Faster, safer releases: ~40% shorter feedback cycles (3 → 5 releases/week) and ~50% faster detection (~58 → ~27 min MTTD) — higher velocity with no rise in escaped-defect risk, critical where patient-facing reliability is non-negotiable.
Defect quality: ~45% higher defect-detection efficiency and ~35% fewer escaped production defects.
Responsible autonomy: Governance built in — decision logging, human checkpoints at release gates, least-privilege access, and protection against prompt injection. Agents handle high-volume analytical work; humans retain authority over strategy and final release decisions.