Work

The work. Marks where we have the right to show them.

Each piece of work is written as problem, approach and result. Logos sit at the top. Where we cannot show a mark, we name the industry instead.

Clients

Marks shown with permission.

Identifiers, not links. Named programmes sit below.

Data engineering and data science

Pipelines that hold. Models the business can run.

The same unit that ships product also builds the warehouse, the feature store and the model that sits on top of them. Data work is not a side deck. It lands in the systems operations already use.

Engineering

Platforms and pipelines

Ingestion from booking, billing, content and device systems. Lakehouse and warehouse design. CDC, quality checks, orchestration and the contracts that keep downstream teams honest.

Science

Models in the product

Demand, pricing, churn, content ranking and anomaly work. Feature stores, evaluation, and a path to take a model offline when it drifts.

Decision

Numbers a meeting can use

Semantic layers, governed metrics and operational dashboards. The same definitions in finance, growth and the floor that runs the product.

Telecom

National DTH / broadband operator

Problem

Hotel, office and apartment partners were running clusters of devices on paper.

Approach

A B2B self-care platform: tagged clusters, room-level packaging, invoices, part and full payments, and temporary suspension.

Result

Enterprise partners could run the estate themselves instead of raising a ticket for every change.

Stack

Relational database · billing events · B2B portal

Travel

Corporate travel platform

Problem

A corporate division serving 750+ companies sat on a third-party application and paid commission on every booking.

Approach

An in-house policy engine and configuration layer, with more than 700 corporate users migrated onto a platform the client owned.

Result

About 20 policy shapes by price, class and designation. The booking path sat inside the client’s own stack.

Stack

Policy engine · web application

Media

National media group - streaming

Problem

Separate applications across news genres were carrying the cost of retention on every stream.

Approach

One front end and back end: home, spotlight, discover, short-form video and personalised stories.

Result

Built to hold more than a million existing downloads on a single product.

Stack

Mobile · content platform

Hospitality

Hospitality network - check-in integrity

Problem

Commission leakage followed the paper register: altered booking IDs, off-platform discounts, no-shows that never hit the platform.

Approach

A digital check-in application bound the stay to a government ID and created an auditable guest record.

Result

The platform gained a penalty path when the register and the booking no longer matched.

Stack

Mobile check-in · identity

Product

ODIO - conversational AI

Problem

Contact centres were scoring a sample of calls and coaching after the fact.

Approach

Every conversation analysed, live assist for the agent, autonomous voice and chat on one stack.

Result

99.9% intent accuracy. 30 to 40% lower AHT. Three times faster resolution.

Stack

Speech · agents · coaching

Product

NaviSlip - marina operations

Problem

Marina operators were running slips, bookings and billing across spreadsheets and whiteboards.

Approach

The management product plus the commercial system around it: site, UX and sales enablement.

Result

Operators and boat owners now share one picture of the marina.

Stack

Operations platform

Education

Int@keOS - higher-education admissions

Problem

Colleges ran paid intake across ads, walk-ins, chatbots and paper files with no shared number for seats or fees.

Approach

One pipeline, an 11-stage counsellor floor, campaign spend tied to fee paid.

Result

Designed for 200 to 20,000 enquiries in a cycle.

Stack

CRM · admissions OS

Product

Zukti - vision AI

Problem

Professional eye measurement still required a chair.

Approach

Digital exams, measurement, fatigue scoring and virtual try-on on the phone already in the user’s hand.

Result

99% accuracy. 50,000+ users.

Stack

Computer vision · mobile

Travel

Bharat Rover - end-to-end travel portal

Problem

The client needed a complete travel product, not a thin wrapper on someone else’s booking engine.

Approach

CRUX teams designed and shipped the portal end to end: flights, hotels, cabs, holiday packages, visa, insurance, eSIM and a 24/7 enterprise desk.

Result

A live consumer and enterprise travel platform at bharatrover.com.

Stack

Travel portal · flights · hotels · packages · enterprise desk

Media · Data

National media group - audience and content platform

Problem

Story, video and app events lived in separate warehouses. Editorial and growth could not share one number for reach, retention or what to publish next.

Approach

A lakehouse for CMS, streaming and app telemetry. Identity resolution across devices. Ranking and personalisation features served back into the product, not only into a dashboard.

Result

One audience picture for news, video and short-form. Editors and the product team used the same definitions for session, completion and return.

Stack

Lakehouse · identity · content ranking · streaming events

Travel · Data

Travel marketplace - booking lake and demand models

Problem

Flights, hotels and packages each kept their own booking truth. Pricing and inventory decisions lagged the market by a day.

Approach

A booking and inventory lake with CDC from the reservation systems, a feature store for route and stay demand, and models for fare and occupancy that wrote back into the pricing path.

Result

Demand and availability sat in one place. Pricing could move on the same day as the market instead of on the next batch.

Stack

CDC · lakehouse · feature store · demand forecasting

Hospitality · Data

Hospitality network - occupancy and leakage analytics

Problem

Check-in, booking and property systems disagreed on occupancy. Commission leakage only appeared after month-end reconciliation.

Approach

An operational warehouse bound stays to the digital check-in record. Anomaly models flagged altered booking IDs, off-platform discounts and no-shows that never hit the platform.

Result

Property and central teams shared one occupancy number during the stay, not after it. Leakage had a path into the same application that ran check-in.

Stack

Operational warehouse · anomaly detection · identity

Consulting · Data

Global consulting firm - reusable analytics workbench

Problem

Engagement teams rebuilt the same extracts, joins and charts for every client. Work product lived in slides and laptops, not in a system the next team could open.

Approach

A governed workbench: reusable data contracts, a metric layer, notebook and dashboard templates, and a path to leave the model with the client when the engagement closed.

Result

New studies started from a shared kit instead of a blank workbook. Clients received a running store, not a file drop.

Stack

Metric layer · notebooks · governed warehouse · delivery kit

iGaming

Sportsbook and casino platforms

Problem

Operators needed specialists who already understand games, sportsbook, wallets, KYC and regulation - not a generalist roster learning the domain on the client’s clock.

Approach

Dedicated CRUX iGaming teams covering casino and RNG games, sportsbook, white-label, mobile, PAM, KYC, payments, reporting and settlement. Online platforms and land-based casino work.

Result

A standing iGaming practice with dedicated teams available against operator roadmaps.

Stack

Sportsbook · casino · wallet · KYC · PAM · settlement