# Manas Labs

> Manas Labs is an applied AI research studio based in Stanmore, Greater London, United Kingdom. It provides AI feasibility studies, focused proofs of concept and model evaluation for defined technical and investment decisions.

The primary website is https://www.manaslabs.co.uk/. Enquiries: hi@manaslabs.co.uk. Shilpa Yadav is Director & Chief Architect.

## Engagements

- [Is the idea worth building?](https://www.manaslabs.co.uk/services/strategic-advisory): Assess an AI idea before funding a build. Manas Labs examines the task, data, alternatives and risks, then delivers a feasibility recommendation and experiment brief.
- [Test the idea on a real task.](https://www.manaslabs.co.uk/services/mvp-prototyping): Test an AI idea through a focused proof of concept. Agree the question, baseline and success criteria, then build a limited experiment with an evidence-led next step.
- [Find out where the answer fails.](https://www.manaslabs.co.uk/services/ai-product-engineering): Evaluate LLM answers and AI models on your actual task. Examine correctness, source support, failure cases, latency and cost with a documented comparison and clear limits.

## Work and company

- [Quantum computing research.](https://www.manaslabs.co.uk/quantum-computing-research): Explore Manas Labs’ early-stage quantum computing research direction: candidate use cases, classical benchmarks and a proposed approach to feasibility studies.
- [A plausible answer is not enough.](https://www.manaslabs.co.uk/work/answer-evaluation-example): Inspect an illustrative answer evaluation: the source, questions, answers and scoring reasons. A downloadable worksheet shows supported, incomplete and unsupported claims.
- [Research with a clear purpose.](https://www.manaslabs.co.uk/about): Manas Labs is an applied AI research studio in Stanmore, Greater London, led by Director and Chief Architect Shilpa Yadav.
- [A question, a test, a reasoned next step.](https://www.manaslabs.co.uk/how-we-work): Our applied research process: frame the decision, design the experiment, examine the evidence and hand over findings with clear limitations.
- [Work you can examine.](https://www.manaslabs.co.uk/case-studies): Explore a worked answer-evaluation example, internal research concepts and reference architecture. Each item clearly distinguishes demonstration from deployed client work.
- [Ideas under investigation.](https://www.manaslabs.co.uk/products): Sentinel, Private RAG, EvidenceDesk and archived ClinicFlow: internal research directions and reference work with explicit status and limitations.
- [Sentinel](https://www.manaslabs.co.uk/sentinel-vault-ai): Sentinel is internal exploratory product work examining laboratory incident support. Read the research questions, proposed evaluation and current limitations.
- [For teams with a question to resolve.](https://www.manaslabs.co.uk/who-we-work-with): Focused AI research for product leaders, technical teams and domain specialists who need evidence before their next investment or product decision.
- [The task comes before the sector.](https://www.manaslabs.co.uk/industries): Applied AI research for evidence-intensive tasks, including finance and nonclinical healthcare operations. Explore realistic use cases and evaluation constraints.
- [Examine the evidence behind the answer.](https://www.manaslabs.co.uk/industries/finance): Explore AI feasibility and evaluation for finance document and internal knowledge tasks, with explicit evidence, data and governance boundaries.
- [Start with a nonclinical question.](https://www.manaslabs.co.uk/industries/healthcare): Research into nonclinical healthcare information tasks: administrative documents, operational knowledge and source-supported review, with candid boundaries.

## Insights

- [What should an AI feasibility study establish?](https://www.manaslabs.co.uk/insights/ai-feasibility-study): Define the decision, review representative data, compare alternatives and design an informative experiment. A practical guide to AI feasibility studies.
- [Define success before building the proof of concept.](https://www.manaslabs.co.uk/insights/ai-proof-of-concept-success-criteria): Define measurable AI proof-of-concept criteria with a clear claim, baseline, held-out examples, failure analysis and explicit proceed, revise or stop decisions.
- [Evaluate the answer, its evidence and its limits.](https://www.manaslabs.co.uk/insights/evaluating-llm-answers): A practical LLM evaluation approach covering correctness, evidence support, completeness, knowing its limits, reviewer calibration and the limitations of model scoring.

## Contact and important context

- [Contact Manas Labs](https://www.manaslabs.co.uk/contact)
- [Privacy policy](https://www.manaslabs.co.uk/privacy)
- [Terms](https://www.manaslabs.co.uk/terms)

Quantum computing status: an exploratory research direction at proposal stage. No developed quantum product, completed experiments, validated results, funding award or confirmed partnership is announced.

Portfolio status: the work listed under products and case studies is internal product development at varying stages, not published client deployments. Sentinel Vault AI is internal product work; Private RAG is a reference architecture; EvidenceDesk is at concept stage; ClinicFlow is archived. The worked answer-review example uses invented source material and authored answers; it is not a client deployment or a model benchmark.