Careem (Uber) · Super App · AI · 2024

Ask
Careem

Took a product-originated brief for an LLM-powered restaurant discovery engine within Careem's Dineout feature and owned the design end-to-end, turning a loyalty benefit into an intelligent recommendations platform that increased average spend per subscriber by 26% and revived the long tail of underperforming restaurants.

Role

Head of Product Design, Pay & Platform

Timeline

Jul 2024 – Sep 2025

Platform Scale

50M+ users

Team

11 designers

+26%
Increase in Dineout spend
+52%
Long-tail restaurant redemptions
44%
Sessions using Ask Careem
Recommendation Engine, Data Flow Architecture
ENGINE TYPELLM-based Context Ranking
LATENCY<300ms p99
A/B VARIANTWinner: +26% AOV
Input Signals (6 Layers)
Order History
Cuisine preferences, frequency, spend patterns, seasonal trends, repeat merchants
Context Signals
Time of day, location, weather, weekday/weekend, nearby restaurants
User Profile
Loyalty tier, dietary restrictions, cuisine blacklist, price sensitivity
Real-time Factors
Active promos, estimated delivery time, restaurant capacity, live ratings
Collaborative Signals
Similar user preferences, trending cuisines, network effects, peer choices
Diversity Scoring
Prevent repetition, explore long-tail merchants, surface new cuisines
Input Normalization
Schema mapping, validation, deduplication
Feature Engineering
Embeddings, normalization, signal weighting
LLM Ranking
Context-aware sort, personality injection
A/B Test Results (4-week Study)
Average Order Value
+26%
Statistically significant (p<0.01)
Long-tail Discovery
+52%
Non-top 100 merchant orders
Cuisine Diversity
+38%
Unique cuisines ordered
Click-through Rate
+34%
Recommendation click rate
Output: Ranked Results
1-5 Recommendations
Ranked by relevance score with 95%+ confidence
Personalized Reasons
Natural language explanations of why (e.g., "Popular with your diner group")
Real-time Indicators
Delivery time, active promos, cuisine tags, distance, ratings
Feedback Loop
Click Events
Which recommendations were engaged, in what order
Conversion Tracking
Order completion → model confidence score mapping
Rejection Signals
Swiped away without action = negative feedback
95%
Recommendation Confidence
3.2s
Avg End-to-end Latency
12M+
Requests per Day
+18%
Session Retention
3.7x
ROI vs. Baseline

Vision.
Alignment.
Execution.

The brief came from product. I took ownership of everything from that point, shaped how the experience would actually work, and orchestrated delivery across engineering, product, and design.

01

Business case and stakeholder alignment, Framed Ask Careem as a retention play: increased Dineout usage = increased Careem Plus stickiness = reduced subscription churn. Secured alignment across three VP-level stakeholders in a single review cycle.

02

Conversational discovery experience, Ask Careem appeared as a persistent entry point, inviting customers to describe their dining needs. The AI responded with 3–5 curated options: restaurant, cuisine, distance, Dineout benefit, and one-tap booking.

03

Long-tail restaurant weighting, Worked with the data team to ensure the AI model weighted lesser-known restaurants more heavily when they matched queries. A design-driven data decision to solve the concentration problem.

04

UI direction and quality bar, Provided hands-on guidance to my Dineout designer, setting the standard for every state: initial prompt, AI processing, results, empty states, errors, and the transition to detail pages.

05

Catalyst for Global AI Search, The Dineout results directly led Careem leadership to ask: "What if we did this for everything?", launching the Global AI Search initiative.

"Ask Careem proved that AI-powered discovery could directly drive commercial metrics in a super app. It wasn't a feature, it was a new way customers related to the platform."

Discovery
reimagined.
Spend
increased.

Within 3 months of launch, Ask Careem had transformed how Careem Plus subscribers discovered dining experiences, with measurable impact on spend, retention, and restaurant partner economics.

Spend Increase

26% increase in average Dineout spend per Careem Plus subscriber within 3 months.

Long-tail Discovery

52% increase in redemptions at previously underperforming restaurants, solving the concentration problem.

Conversion

39% of Ask Careem queries converted to a restaurant visit within 48 hours.

Retention

+8% increase in Careem Plus renewal rate among active Dineout users.

Next Case Study

Careem:
Global AI Search

Want in?

Score 100% on the Design Eye test to earn access. Or enter the password if you have it.

4 rounds. No partial credit.