Ruchi Parijat
Work Approach Story Writing Contact

Philosophy

How I Work

Three principles — and the concrete processes behind them.

Principle 01

Platform Before Features

I don’t build point solutions. I build the shared foundation that makes every downstream experience possible — governed, observable, and designed to scale across teams.

A feature that can’t be replicated across the platform isn’t a product win — it’s technical debt wearing a product hat. At Deloitte, this meant designing the data unification layer before a single AI feature was shipped.

In Practice: When a Stakeholder Pushes for a Feature First

I run a “replicate or retire” exercise: can this feature be extended to 3 other teams within 90 days without re-engineering? If no, it goes on the platform backlog, not the product roadmap.

When I’ve had to say no to a feature request, I come back with a platform design that makes that feature trivially easy to build — usually within the same quarter.

Principle 02

Compliance Is a Day-One Requirement

In regulated industries, compliance embedded after the fact is not compliance — it’s a liability. KYC/AML filters, audit trails, and access controls belong in the architecture, not the sprint backlog.

Every retrieval pipeline I’ve built has compliance logic at the data layer, not the application layer.

In Practice: How I Run Discovery in Regulated Environments

Before writing a single user story, I map the regulatory surface: which data fields are PII, which workflows trigger audit events, and which decisions require human sign-off by law. This becomes a “compliance contract” signed off in week one.

On the Deloitte AI Assist project, this upfront compliance mapping cut legal review cycles from 6 weeks to 10 days.

Principle 03

Data Earns Trust Before AI Spends It

Every AI system is only as reliable as the data it runs on. I’ve seen teams build impressive demos on top of stale, duplicated, or poorly governed data — and watched those demos collapse in production.

My approach: data quality, lineage, and governance come first. AI capabilities come second.

In Practice: My Data Readiness Framework

Before any AI feature goes to staging, I run a four-gate check: freshness, completeness, lineage, and access control. All four must be green before AI features are enabled in production.

This framework came from a near-miss where a retrieval pipeline was returning outdated fund performance data with high confidence scores. I shipped the four-gate check as an automated pre-deployment test the same week.

© 2026 Ruchi Parijat