Kaushal Khodifad is an AI product leader for regulated industries and the founder of CLOZOM, the professional network for India's healthcare professionals. Eight years shipping product, the last four with models in the loop, where compliance, audit trails, latency budgets and real users decide what reaches production. PGDABA, IIM Ahmedabad. The production record is at /work, the working AI products at /products, and contact at /contact.
AI product leader for regulated industries.
17 working AI products across 9 industries.
The model isthe easy part.Shipping it inside compliance, audit trails, latency budgets and real users is the hard part.
Eight years shipping product, the last four with models in the loop. The industry changes; the hard part does not.
LinkedIn · 10K+ followersProduction outcomes by Kaushal Khodifad: at Infosys (2018 to 2021), credit decisioning for Fortune 500 banking clients that took loan processing from seven days to under 24 hours, under SOX and GDPR; at Dukaan (2022), ML carrier optimization saving $1.5M+ a year across 5M+ merchants; at OTPless (2023), a 60% lift in authentication success in eight weeks on a platform serving 30M+ end users; for a $200M+ manufacturer, demand and inventory forecasting in production that improved forecast accuracy 40% and cut inventory carrying costs 25%; and CLOZOM, the professional network for India's healthcare professionals, which he founded in 2023 and has grown to 50,000+ registered professionals and 26,000+ jobs. Full record at /work.
- 50,000+Registered healthcare prosCLOZOM
- +60%Authentication successEnterprise SaaS platform
- $1.5M+Annualized savingsVenture-backed e-commerce
- <24hLoan processing, down from 7 daysFortune 500 banking clients
- 25%Lower inventory carrying costs$200M+ manufacturer
Selected working AI products by Kaushal Khodifad: Arbiter (Insurance & Lending): Commercial credit · every fact quoted, every ratio and gate in code; Arintra (Healthcare & Life Sciences): Autonomous medical coding · ICD-10 and CPT; Finance Copilot (Finance & Accounting): CFO-office answers on runway, burn and AR · the model narrates, the engine computes; Project Atlas (Banking & Payments): Chase card data · governed access, natural-language queries, an audit trail; Praxis (AI Platform & Compliance): Agent control plane · every action policy-checked, ledgered and human-approvable; Azeon (Sales, Marketing & Service): Customer-service agents that act · four policy gates and an evidence ledger. Independent builds on public and synthetic data: not client work, and not affiliated with or endorsed by any company named. All 17 are at /products.
Where the controls live.
Each product takes a regulated decision and puts the control in code you can open, not on a slide.
17 working products9 industries
- Insurance & Lending
Arbiter
Commercial credit · every fact quoted, every ratio and gate in code
Open the product - Healthcare & Life Sciences
Arintra
Autonomous medical coding · ICD-10 and CPT
Open the product - Finance & Accounting
Finance Copilot
CFO-office answers on runway, burn and AR · the model narrates, the engine computes
Open the product - Banking & Payments
Project Atlas
Chase card data · governed access, natural-language queries, an audit trail
Open the product - AI Platform & Compliance
Praxis
Agent control plane · every action policy-checked, ledgered and human-approvable
Open the product - Sales, Marketing & Service
Azeon
Customer-service agents that act · four policy gates and an evidence ledger
Open the product
Independent builds on public and synthetic data. Where a real company is named, it anchors the domain only: none of this is client work, and none of it is affiliated with or endorsed by any company named.
Teardowns, not takes.
Long-form on how AI gets into production inside regulated industries. Every claim carries a primary source.
- Grounding is the easy half of autonomous medical coding - an Arintra teardownValidating every code against the actual CMS tables stops a model inventing codes, and that fight is largely won. It does not stop a real row from being the wrong answer, which is the failure mode nobody demos.20 min read
- How card-data entitlements actually work - an Atlas teardownAn entitlement that lives in application code is a claim about every code path that will ever exist. An entitlement that lives in the database is an artifact somebody else can read.19 min read
- Compliance evidence: automating the part auditors actually check - DPDP Rules 2025 teardownA regulator has already said it plainly: pointing at a correctly configured system is not proof that it was configured correctly on the day it mattered. Compliance tooling automates the configuring, and the proving is still a person with a folder of screenshots.20 min read
- How a lending decision gets warranted, not just made - a credit memo automation teardownMaking the number is the easy half. Warranting it, with reason codes a regulator accepts and a trace nobody can quietly edit, is the half that decides whether the system ships.19 min read
- What actually stops an agent is code it cannot reach - tearing down two agent prototypesEvery governance feature that lives inside the model's context window is advisory. The test for any agent control is one question: can the model change this outcome by saying something?19 min read
Contact Kaushal Khodifad, AI product leader for regulated industries, by email at kaushal@kaushalkhodifad.com or via LinkedIn. Based in Bengaluru, India.
Does your AI have to pass an audit?
Write to me directly. I read every note myself.