# Saurav Patro

Product leader. Marketplaces, partner platforms, incentive design, commercial workflows and LLM products.
Based in Dubai, UAE. Open to relocation.
Contact: saurav.patr@gmail.com · LinkedIn: https://linkedin.com/in/saurav-patro-a2164330

## Summary

Ten years building marketplaces from the inside, on every side: the platform side at Careem, the seller side at Amazon, the delivery-partner side at Zomato, a three-sided travel marketplace at Travel Triangle, a two-currency game economy at Hike, and B2B logistics billing at Delhivery. Builds the rails, the incentives on top of them, and the measurement that shows whether each side is winning. Ships LLM products and builds with Claude Code.

## Roles

| Dates | Company | Title |
|---|---|---|
| Mar 2024 to present | Careem (MENA super-app, 50M+ users) | Lead Product Manager, New Bets (AI & Enterprise) |
| Apr 2023 to Jan 2024 | Zealth.club | Chief Product Officer & Co-Founder |
| Apr 2022 to Mar 2023 | Hike | VP, Chief of Staff to CEO |
| Dec 2020 to Apr 2022 | Amazon | Product Leader, Seller Growth & Quick Commerce |
| Sep 2018 to Dec 2020 | Zomato | Group Product Manager, Online Ordering |
| Jun 2017 to Jul 2018 | Travel Triangle | Product Leader, Customer Experience |
| Dec 2015 to Jun 2017 | Delhivery | Manager, Business Operations (among the first 30 employees) |

Education: MBA, XLRI Jamshedpur (2013 to 2015, top 5%). BBA, Central University of Jharkhand (2009 to 2012, top 1%); founded Unnayan, free tuition for 7,000+ village students.

## Case studies

### Careem Tenants (partner platform)
- Third-party businesses plug into Careem Pay, discovery, the design system and the customer funnel; fulfilment stays with the partner. Vetting and agreed commercials gate listing.
- 20+ partners in Dubai, 15 in Jordan, 7 in Egypt within six months.
- Careem grew from about 5% of partner business at launch to about 96% of a car-rental partner's GTV and about 92% of a home-services partner's.
- Customers who adopted a partner service tripled app spend (45%+ of adopters).
- Lasting partnerships include Justlife (home services) and Washmen (laundry).
- Car-rental partner Swapp: inherited a strained partnership (live since 2022), rebuilt the joint roadmap within six months, proved the new model with a proof of concept, then closed an exclusivity deal with equity for Careem, its first stake in a partner app (publicly announced 2025).

### Careem for Business (B2B, zero to one)
- Signal: month-end support tickets for tax invoices (agent-only tickets rose to about 70% in the last five days of the month).
- Research: 15+ users across three markets; employer reimbursement of 20 to 40 trips and orders a month; real-estate agents as a key segment.
- Validation: fake-door page built on Lovable; 200+ interest forms from about 70 companies in two weeks.
- Product: admin workspace with per-service spend policies and employee lists; company-funded preloaded wallet with daily-expiring allowances; "pay as business" at checkout.
- Go to market: sold the first accounts personally; $200K a month in revenue three months after launch; then handed to a sales team and the commercial org.
- North star: new business accounts; revenue as output.

### AI grocery list (LLM product)
- Photo, pasted WhatsApp list, or voice agent. V1 used human transcribers as the baseline; the LLM shipped after matching it.
- 40-click weekly shop to about 3 clicks. 22% week-one adoption against a 5 to 7% target; 65% repeat within two weeks; adopter order value about 70 to about 250 AED.
- Refused a big-basket discount; used country-of-origin tags to anchor quality.

### AI Meals on Careem Food (personalisation and habit)
- Diagnosis: the food funnel lost about 30% from search to cuisine, 50% from cuisine to restaurant and 40% from menu to checkout; office-goers ordered from the same 10 to 12 nearby restaurants.
- Product: per-user carousel of complete set menus, a location layer surfacing best sellers popular in the user's building, single-tap checkout.
- Piloted in Abu Dhabi only, Dubai held back until repeat was proven; two teams pushed back over quarterly order-value targets and were aligned with a prototype and a repeat commitment.
- Result: ~60% three-day repeat among users ordering 10+ times a month against a 40% target set before launch; order value fell as planned; gross profit per user +8%; higher conversion to Careem Plus.

### AI Enablement (Careem)
- Ask Careem: retrieval assistant over Confluence, Notion and Drive in Slack, raising Jira tickets; actions validated on a shadow replica of production; about 98% accuracy after rollout.
- Up to 80% of commercial and operations workflows run through AI; merchant onboarding 6 days to 1; merchant revenue +24% from AI merchant tooling and dynamic pricing.
- Published an analysis skill on Careem's internal skill marketplace.

### Zealth (co-founder and CPO, health-records network)
- Consent-based health-records layer on India's ABHA rails; one of the first five companies admitted to the government sandbox.
- Five partner groups: government (ABHA sandbox), insurers, hospital chains (8 hospitals in the pilot), diagnostic labs (reports delivered into the record), private clinics (40 in the pilot).
- Products: longitudinal record with time-limited consent for patients; patient-history and reach-out tool for doctors and labs.
- Pilot earned revenue; shut down when the government deprioritised ABHA.

### Hike (game economy)
- Two-currency economy; pay-to-play loop converting 8% of new users to paid monthly; Rush VIP tiers raised spend 30% and D30 retention from 12% to 30%.
- $100M in recurring revenue, profitable from day one; led a 230-person team through the pivot.

### Amazon (seller side)
- Mobile-first seller onboarding for micro-sellers behind 80% of India volume; $130M incremental revenue.
- Top Brands trust signal: D2C brand revenue +200% at 3% higher conversion. Quality engine on 30+ signals cut returns 25%.

### Zomato (fulfilment engine and incentives)
- North star "shutdown" (time the app showed nothing deliverable) driven toward zero.
- Built a 90K-cyclist fleet from zero: 43% of orders at 50% lower cost; cost per order about ₹120 to about ₹78.
- Rebuilt assignment (multimodal), partner profiles, payouts; guardrails for women riders and riders with disabilities; first delivery-partner insurance.
- Payouts rebuilt in six phases, ending with cohort-level rate cards for groups as small as 10 riders; part-time rate cards.
- First promotion platform: automatic offer stacking in one cart and a targetable platform-fee structure.
- Karma Score fraud system cut abuse 90%; dynamic pricing improved margins 18%; 250-city P&L from -20% to +16% EBITDA.

### Travel Triangle (three-sided marketplace)
- Travellers, local travel agents and on-ground providers. Repeat was about 5% over nearly five years.
- Referral reward mechanism and loyalty rewards; north star of trips per converted traveller (about 20 per conversion).
- Weekend getaways drew 90% of repeat travellers and 40% of revenue within six months.
- Lead scoring on 25+ attributes improved lead-to-conversion 23%.

### Delhivery
- DMiles prepaid wallet for long-tail B2B shippers: $3M unrealised revenue recovered, leakage -95%.
- AddFix address-resolution engine, later offered to Amazon as a service.

## Built personally
- Cubby: baby-care app on Firebase with per-baby access controls.
- Journey Log: fitness web app on Cloudflare Pages.
- Claude skills: Product Board venture review, competitive app teardown.

## Full record of delivered work

### Careem

- **Tenants partner platform** (2024 to now): Vetted third-party businesses plug into Careem Pay, discovery, the design system and the customer funnel; fulfilment stays with the partner, and agreed commercials gate listing. Results: 20+ partners in Dubai, 15 in Jordan, 7 in Egypt in six months; Partner GTV via Careem ~5% to ~96% (car rental), ~92% (home services); 45%+ of adopters tripled their app spend.
- **Swapp partnership turnaround** (2024 to now): Took over a strained car-rental partnership live since 2022, rebuilt the joint roadmap within six months, proved the new model with a proof of concept, then closed an exclusivity deal. Results: Exclusivity, with equity in Swapp for Careem: its first stake in a partner app.
- **Justlife and Washmen partnerships** (2024 to now): Long-term partners in home services and laundry, onboarded onto Tenants and grown with the platform.
- **Careem for Business: the discovery** (2024 to now): While running a six-month care-automation strategy, ticket data showed agent-only queries rising from about 10% on a normal day to about 70% in the last five days of each month, driven by tax-invoice requests. Careem issued transaction receipts, not tax invoices. Interviews with at least 15 users across three markets showed employer reimbursement of 20 to 40 trips and orders a month; real-estate agents were the surprising segment. Results: Same pattern across UAE, KSA and Jordan.
- **Careem for Business: fake-door validation** (2024 to now): A fake-door site built on Lovable and shown through a small checkout banner only to the cohorts that had asked for invoices. Unexpected demand surfaced for intra-city document and parcel delivery, which companies were doing by booking taxis. Results: 200+ interest forms from ~70 companies in two weeks; Most companies 100 to 400 staff, with a tail from 50 to 10,000; Typically only 20 to 30 employees per company already on Careem.
- **Careem for Business: the product** (2024 to now): A company-funded closed wallet on Careem Pay rails with daily allowances that expire at end of day, per-service spend policies, an admin web portal, and pay-as-business at checkout. Funded through a proposal to Careem's leadership team about a month after the pattern appeared, while still delivering the care-automation goals. Results: North star: new business accounts; revenue treated as an output; First accounts sold personally.
- **Careem Plus membership** (2024 to now): Owned the paid membership programme end to end and rebuilt onboarding, cross-sell and upsell. Results: Trial-to-paid 40% to 70%; Retention 40% to 65%.
- **AI Meals on Careem Food** (2024 to now): Complete set meals from nearby favourites with one-tap checkout, plus a location layer surfacing what is popular in your building. Piloted in Abu Dhabi only, with a target set before launch. Results: ~60% three-day repeat vs a 40% target (users ordering 10+ times a month); Gross profit per user +8% while order value fell; Higher conversion to Careem Plus.
- **AI grocery list** (2024 to now): Photograph a list, paste one forwarded on WhatsApp, or speak to an agent; matched against purchase history and inventory into a single-page basket. Version one ran on human transcribers as the baseline the model had to beat. Image input first, voice later, sequenced on cost. Results: 40-click weekly shop to about 3; 22% week-one adoption vs a 5 to 7% target; 25 to 30% sustained; 65% of adopters repeat within two weeks; Adopter order value ~70 to ~250 AED; Uber and noon shipped equivalents within about two months.
- **Grocery revamp, five workstreams** (2024 to now): A research-led programme run in parallel: order-volume uplift, price perception (quick commerce is not more expensive than local retail), a fresh-produce quality pivot, country-of-origin tags on SKUs, and a full design overhaul. The shopping list was the hardest to get approved because it was new to the market. Results: Refused a big-basket discount; country-of-origin tags anchored quality instead.
- **Hypermarket research method** (2024 to now): In-person intercepts inside hypermarkets with shoppers above a spend threshold, asking how they prepared for the trip rather than what they bought, with live price comparisons. Found a sub-1% cohort already building week-long carts by hand; packaging waste surfaced as a motivation nobody had expected. Results: Saturation after the sixth or seventh user.
- **People also buy** (2024 to now): Users finished the list and still browsed the aisles, which falsified the team's hypothesis. The response was a checkout strip surfacing forgotten items from the previous cart and complementary products.
- **Company-wide AI hackathon** (2024 to now): Created and funded a hackathon run as an operating mechanism: every vertical (rides, groceries, food, car rental, micro-mobility, payments) submitted its own top problems against the company's top four goals. My team then took the one goal every vertical had declined: raising grocery order value.
- **AI Enablement charter** (2024 to now): Pitched and founded the charter as three people; it became Careem's central AI and platform function and the implementation partner for every vertical across design, engineering, PRDs and data. The team's tooling moved from ChatGPT to Gemini to Claude Code. Results: Up to 80% of commercial and operations workflows run through AI; From founder and board level down to ops executives.
- **Ask Careem** (2024 to now): A retrieval assistant over Confluence, Notion and Drive process docs, in Slack, that raises Jira tickets and runs workflows in one click. Agent actions were validated on a shadow replica of production before rollout; it regenerates its own documentation whenever features ship. Results: ~98% accuracy after rollout, with thumbs up and down feedback.
- **Merchant onboarding AI** (2024 to now): Menu ranking and clustering, image-quality enhancement and dish classification (including veg and non-veg) for restaurant onboarding. Results: Merchant onboarding from 6 days to 1.
- **AI merchant tooling and dynamic pricing** (2024 to now): AI tools for merchants on the platform, including dynamic pricing. Results: Merchant revenue +24%.
- **Restaurant and merchant platform rebuild** (2024 to now): Rebuilt after benchmarking against Keeta and DoorDash, redesigned for four personas: owner, manager, restaurant staff and kitchen staff.
- **Super Portal** (2024 to now): One internal platform for promotions, inventory, cataloguing, merchant onboarding, quality checks and complaint handling, with role-based access. Design rule: the newest employee should reach anything in five clicks.
- **Selection intelligence** (2024 to now): Ran store-level selection for grocery, and merchant and dish availability for Careem Food benchmarked against other platforms, end to end: signals, matching, gap ranking, routing to action and measurement.
- **Dark-store layout optimisation** (2024 to now): Work-time studies and computer vision to lay out fulfilment-centre zones by demand pattern, with store-level specialisation across a cluster. Results: Layouts refreshed about every six months.
- **Care automation strategy** (2024 to now): A six-month strategy with two PMs: one on refunds, incorrect fares and fees, with direct monetary impact; one building a full FAQ corpus to power a later retrieval system.
- **Internal agents: bug investigator, research agent and the RCS skill** (2024 to now): An agent that handles most bug triage, with humans only confirming which step in a flow broke; an internal agent with context across PRDs, specs and databases that turned a multi-day, five-person report chase into a sourced one-pager; and an analysis skill (cross-analysis, dependencies, leakage detection) published on Careem's internal skill marketplace and used across the company.
- **Team, leadership and mentorship** (2024 to now): Seven direct reports, five product managers and two APMs, with engineering and design working to the charter on a dotted line, across five verticals: rides, food, grocery, enterprise and internal platforms. The AI Enablement charter started as three people and Careem for Business was built by a six-person team. The design intern on that first three-person team is now a design manager, and the engineer who volunteered onto the first grocery-list team now runs Careem's web experience. Results: 7 direct reports: 5 PMs, 2 APMs; Nimble teams of 3 and 6 that built new lines.

### Zealth

- **Consent-based health-records layer** (2023 to 2024): As co-founder and CPO, built a layer on India's ABHA rails that normalised imaging, pathology and lab reports into one longitudinal record, shared through time-limited consent. A patient app, plus a history and outreach tool for doctors and labs. Results: One of the first five companies admitted to the government sandbox.
- **Five partner groups on one record** (2023 to 2024): Government (ABHA sandbox), insurers, hospital chains, diagnostic labs and private clinics. Results: Pilot across 8 hospitals and 40 clinics; The pilot earned revenue.
- **Shutdown** (2023 to 2024): Closed when the Government of India deprioritised ABHA, the health-records mission it ran on, alongside resistance from hospital incumbents.

### Hike

- **Messaging to gaming pivot** (2022 to 2023): As VP and Chief of Staff to the CEO, ran the company internally and led the pivot from messaging to real-money gaming. Piloted five game formats (web3, non-web3, cash, non-cash, freemium) against each other before committing to one. Results: 230-person team through the pivot in 6 months; $100M in recurring revenue, profitable from day one.
- **Two-currency game economy** (2022 to 2023): Ran the full economy of Rush's two-currency ecosystem and built the pay-to-play loop for the freemium tier. Results: 8% of new users converted to paid every month.
- **Rush VIP tiers** (2022 to 2023): A tiered, gamified loyalty programme. Results: User spend +30%; D30 retention 12% to 30%.
- **Data and Revenue team** (2022 to 2023): Set up Hike's first Data and Revenue team and made A/B testing and pricing governance the default path to ship. Results: Margins +20%.

### Amazon

- **Seller onboarding and gamification** (2020 to 2022): Owned the seller gamification product and redesigned onboarding mobile-first for the micro-sellers behind 80% of India volume, many of whom had relied on cyber cafes because Seller Central was web-first. Results: $130M incremental revenue.
- **Top Brands trust signal** (2020 to 2022): A product-page trust signal for good D2C brands, built after work on counterfeit and bad-seller detection. Results: D2C brand revenue +200%; 3% higher conversion.
- **Seller quality engine** (2020 to 2022): An ML engine on 30+ signals that blocked fraudulent sellers and items before they reached customers. Results: Returns -25%.
- **India Ultrafast Fresh** (2020 to 2022): Founding team; unified inventory across warehouses. Results: Daily orders +20%; Fulfilment cost -10%.

### Zomato

- **Cyclist fleet** (2018 to 2020): Order data showed 80 to 85% of orders came from within about 3 km, so short trips went to cyclists, unlocking supply from milk and newspaper delivery riders who sat idle at lunch and dinner. Piloted in Delhi, Gurgaon, Noida and Hyderabad, then rolled out nationally. Results: 90K cyclists from zero; 43% of orders at 50% lower cost; Cost per order ~₹120 to ~₹78.
- **Rider safety and insurance** (2018 to 2020): No underwriter would cover cyclists, so I introduced delivery-partner insurance at Zomato for the first time. Built guardrails into assignment: no night orders for women riders, no multi-floor buildings for riders with disabilities. Results: Zomato's first delivery-partner insurance.
- **The hidden cyclist-or-biker choice** (2018 to 2020): Pilots showed that showing customers a cyclist-or-biker choice raised complaints even though measured ETAs differed by only 30 to 45 seconds, so the choice stayed hidden. A fast-follow made assignment concentric: cyclists first close in, bikers otherwise.
- **The "shutdown" north star** (2018 to 2020): Defined the north star as "shutdown", the share of time the app showed nothing deliverable in an area, with a goal of zero rather than a reduction.
- **Payouts and rate cards** (2018 to 2020): Rebuilt rider payouts in six phases with the finance org: multimodal payouts, then incentive tiers by partner type, then cohort-level rate cards and part-time rate cards. Results: Rate cards for cohorts as small as 10 riders.
- **Promotion and fee platform** (2018 to 2020): Zomato's first promotion platform: bank, wallet, partner and Gold offers stacked automatically in one cart, plus a platform-fee structure that could be added, removed and targeted by cohort.
- **Karma Score and dynamic pricing** (2018 to 2020): An ML fraud and abuse system, alongside a dynamic pricing engine. Results: Platform abuse -90%; Margins +18%.
- **Fulfilment pod and P&L** (2018 to 2020): Led the fulfilment product pod, with APMs and program managers across assignment, serviceability, payouts, and fraud and cash collection, through growth from 5 to 500 cities. Results: 250-city P&L from -20% to +16% EBITDA on a $20M budget.

### Travel Triangle

- **Weekend getaways** (2017 to 2018): Domestic weekend-getaway packages that turned a once-in-a-lifetime purchase into a repeat one, in a business where repeat had been about 5%. Results: 90% of repeat travellers within six months; 40% of revenue within six months.
- **Referral and loyalty loop** (2017 to 2018): Referral rewards and loyalty rewards across a three-sided marketplace of travellers, local travel agents and on-ground providers, with trips per converted traveller as the north star.
- **Lead scoring and conversion engine** (2017 to 2018): A retargeting and conversion engine that scored and re-engaged leads on user profile, location and 25+ profile attributes; led a 40-person CX team. Results: Lead-to-conversion +23%.

### Delhivery

- **DMiles prepaid wallet** (2015 to 2017): Moved risky long-tail B2B shippers from credit to a prepaid wallet. Results: $3M unrealised revenue recovered; Leakage -95%.
- **AddFix address resolution** (2015 to 2017): Indian addresses are free text, so one door arrives written many ways. AddFix clustered variants into uniform locations by learning from successful deliveries, delivery alternates and shared markers, giving routing and capacity planning a structured entity. Later offered to Amazon as a service. Results: Among Delhivery's first 30 employees.

### Education

- **MBA, XLRI Jamshedpur** (2013 to 2015): International live projects and exchange in the USA and China. Results: Top 5% of batch.
- **BBA, Central University of Jharkhand** (2009 to 2012): Strategy and Marketing. President of India Scholarship. Founded Unnayan, which gave village students free tuition from student volunteers. Results: Top 1% of class; 7,000+ students reached through Unnayan.