When to hire an AI consultant
- You have an AI feature on the roadmap but no one has tested it on your real data.
- A prototype works in a demo and fails on messy documents or edge cases.
- You handle personal or health data and need to plan for DPDP Act 2023, GDPR or HIPAA-style obligations.
- You are choosing between a vendor API, an open model you host, or a custom pipeline.
Engagement types
- AI readiness review (two weeks, fixed scope): data inventory, use-case ranking, risk review and a prioritized roadmap.
- Architecture audit: a review of the existing system and its LLM integration points.
- Fractional advisory retainer (10–20 hours a month): architectural steering, hiring evaluation and technology selection. See the advisory packages.
Decisions I help with
- Build vs buy, and model hosting: cloud API, or local and hybrid deployment.
- Hallucination control with schema-constrained output and deterministic validation.
- Personal data and consent under the DPDP Act 2023 and GDPR.
- Human review for the exceptions that validation cannot resolve.
Proof
The approach comes from production work: schema-constrained extraction pipelines that reach over 99.4% field-level precision, and the iCare platform, deployed across state hospital networks. Read the essay on structured data extraction with LLMs or the iCare case study.