LLM & RAG evaluation
Find hallucinations, retrieval failures and regressions before they reach users.
- Evaluation datasets and quality rubrics
- Grounding, relevance and accuracy checks
- Prompt and model regression analysis
BIRMINGHAM, UK · INDEPENDENT CONSULTING
Turn uncertain AI behaviour into clear, reviewable evidence. I help teams evaluate LLMs, validate RAG systems, and build automation they can trust.
AI Quality Architect · Test Architect · Senior SDET
14+ yearsEnterprise quality engineering
Healthcare & financeDomain-aware testing
Evidence firstExplicit criteria, visible limitations
WHERE I CAN HELP
Start with a focused scope. Agree the evidence needed before choosing the tools.
Find hallucinations, retrieval failures and regressions before they reach users.
Build maintainable checks for the journeys that matter most to your product.
Connect test results, human review and observability into a clear release decision.
PUBLIC ENGINEERING EVIDENCE
Selected public reference projects. These demonstrate engineering approaches using synthetic or controlled examples; they are separate from employment outcomes.
A provider-neutral approach to datasets, deterministic metrics, regression comparisons and explicit release policies. The reference mode runs offline with synthetic data.
Typed UI/API tests, fixtures, cross-browser configuration, diagnostics and configurable quality gates. A documented local API path gives reviewers a focused starting point.
THE EXPERIENCE BEHIND THE WORK
I’m Ashok Kumar Manohar, an AI Quality Engineering Architect with experience at Happiest Minds Technologies, Acentra Health and Wipro. My work spans test architecture, healthcare API quality, banking regression and AI evaluation.
My approach combines deterministic checks, calibrated evaluation and clear human accountability. I favour practical systems that another engineer can inspect and maintain.
Professional background ↗SELECTED EMPLOYMENT OUTCOMES
LET’S DEFINE THE FIRST MILESTONE
Share your application, AI workflow or test coverage challenge. We can agree a focused scope, acceptance criteria and handover.