Executive Portfolio · AI Product & Innovation Leadership

I Build AI Products That Ship.

15+ years turning frontier technology into production revenue — as a founder, operator, and product executive across Disney, Starbucks, HP, FedEx, National Grid, and the GCC's most ambitious platforms.

Years in product & innovation

Global projects delivered

$0M

Venture portfolio managed

Markets — India · UAE · KSA · Japan · Singapore

Explore the Vision Jump to the 12 Use Cases

40% conversion lift at Disney Parks$7M incremental revenue at HP25% barista availability gain at StarbucksAgentic copilot piloted with National Grid ESO$20M+ innovation pipeline builtAcquisition exit 2019

The Problem I Solve

Most enterprise AI never ships. Mine does.

Analysts estimate the majority of enterprise AI initiatives stall between pilot and production. The gap is rarely the model — it's product judgment: scoping what's buildable, proving value before engineering spend, and owning performance after go-live. That gap is exactly where I operate.

From boardroom to build.

I prototype in code with frontier models before a single engineering sprint is committed — compressing discovery-to-build cycles and de-risking scope.

From pilot to production.

I define the evals, guardrails, and human-override logic that turn a demo into a system enterprises trust — and I stay accountable after deployment.

From product to P&L.

Founder and portfolio operator: every product decision I make is framed in revenue, cost, and risk — the language of the executive team.

Operating Philosophy

Human-centered. Data-driven. Production-obsessed.

Human-Centered Innovation

Every system starts with the person using it. I design for adoption, not demos: if operators don't trust it, it doesn't ship.

Data Decides

Model selection, feature priority, pricing — every call is benchmarked and measured. Conviction is cheap; evidence compounds.

Prototype Before Promise

I build working proofs-of-concept myself before committing teams. Stakeholders experience the product before they fund it.

Adapt and Pivot

From e-commerce to edge AI to agentic systems: I've re-platformed my expertise through every technology wave and taught teams to do the same.

How an engagement runs

1Diagnose
2Prototype
3Validate
4Ship
5Own Performance

Capabilities

Breadth where it matters. Depth where it counts.

AI & Technical

Agentic AI SystemsMulti-Agent ArchitectureLLM Product DesignPrompt Engineering & EvalsFrontier Model PrototypingComputer VisionEdge AI & Industrial IoTDigital Twin PlatformsPredictive AnalyticsVoice AI & ASR

Product & Strategy

0→1 Product DevelopmentEnterprise B2B SaaSPlatform ArchitectureGo-to-Market StrategyP&L OwnershipInnovation StrategyRapid Prototyping (code & no-code)Squad LeadershipCRM Integration

Markets & Industries

GCC · UAE · Saudi ArabiaSouth East AsiaIndiaEnergy & UtilitiesAutomotiveAviationMaritime & IndustrialHealthcareRetail & E-commerceTelecommunicationsTravel & Hospitality

Track Record

Trusted by category leaders.

Eighteen brands across five continents — from Fortune 100 icons to sovereign-scale platforms. Six engagements below, told the way I ran them: problem, approach, outcome.

Walt Disney

Starbucks

FedEx

Expedia

National Grid ESO

NATS

KAUST

Maersk

Ma'aden

Mitchell International

Jetour · Elite Group

Backroads

Grover

Circles.Life

Celcom Axiata

Novant Health

Armada AI

01Agentic AI · Automotive Retail · GCC

Jetour Automotive AI Agent Platform

Elite Group Holding, UAE

Problem: The showroom-to-ownership journey restarted at every touchpoint — customers repeated themselves, sales lacked context, aftersales leaked retention.
Approach: Designed a four-agent system on the NEXUS platform — Sales & Discovery, Test Drive Booking, Service & Aftersales, Loyalty & Retention — orchestrated by a central reasoning layer over live vehicle catalogue, DMS pricing, CRM, and showroom availability. Built the interactive proof-of-concept personally, with bilingual Arabic/English voice architecture and Salesforce handoff design. Core insight: “the journey never restarts.”
Outcome: Full pre-sales win — executive briefing, demo runbook, persona journeys, and a Gulf expansion plan across UAE, KSA, and Qatar.

4 orchestrated agents
Bilingual voice architecture
3-market Gulf expansion plan

02Agentic AI · Energy · UK Regulated

NEXUS Helm — AI Copilot for Grid Operations

National Grid ESO

Problem: Grid operators lost critical minutes reconstructing queries across multiple systems during live events.
Approach: Designed Helm, an agentic copilot that answers natural-language questions against live asset data — owning the prompt framework, tool-calling schema, context strategy, and hallucination guardrails, benchmarked against domain experts.
Outcome: Live pilot with National Grid ESO; extended to NATS (aviation) and KAUST (ecology). Positioned upstream of the UK Data Sharing Infrastructure; SIF/NIA innovation funding submissions led with NGT, NESO, and Cadent.

3 sector deployments
Expert-benchmarked evals
UK DSI positioning

03Edge AI · Industrial · $100M+ raised

Agentic Edge Intelligence

Armada AI — Maersk · Ma'aden · Tampnet

Problem: Maritime, mining, and energy operations generate telemetry faster than cloud round-trips can act on it.
Approach: Designed the product framework for AI agents running on Galleon mobile data centres — LLM-assisted anomaly summarisation for vessel fleets, computer-vision safety monitoring, and quantised-model evaluation for offline edge deployment.
Outcome: Multi-million-dollar Galleon deals secured and a strategic GCC partnership led — capability translated directly into enterprise revenue.

Multi-million $ deals
Strategic GCC partnership
Offline-capable agents

04Agentic AI · Retail Operations

Agentic Shift Optimisation

Starbucks

Problem: Managers spent hours hand-building schedules; peaks ran understaffed while quiet hours burned margin.
Approach: Shipped an agent that processes live traffic, promo calendars, and skill matrices to generate optimised schedules — with confidence thresholds and manager-approval override built in.
Outcome: Deployed in production with sustained gains across pilot stores.

25% barista availability gain
15% store revenue lift
6+ manager-hours reclaimed weekly

05E-commerce · Personalisation · Team of 13

Disney Parks E-commerce Transformation

The Walt Disney Company

Problem: Rich first-party guest data, generic experience — conversion lagged benchmarks.
Approach: Led 13 engineers and designers through a platform redesign built on a rigorous experimentation framework: personalised recommendations, journey redesign, validated rollout.
Outcome: The experimentation framework kept compounding performance after the engagement ended.

40% conversion lift
Team of 13 led
Framework outlived engagement

06LLM · Insurance · Human-in-the-Loop

LLM Claims Intelligence

Mitchell International

Problem: Manual first-pass claims review throttled the entire adjudication pipeline.
Approach: Led an LLM system that triages claims, flags anomalies and missing documentation, and prioritises adjuster queues — with confidence thresholds and escalation rules evaluated against adjuster benchmarks pre-launch.
Outcome: First-pass manual review eliminated; adjusters refocused on complex judgment calls.

30% faster processing
First-pass review eliminated
Benchmarked pre-launch

HP — $7M incremental annual US revenue via AI personalisation (4% uplift)

FedEx — Sustainability Index SaaS — multi-million-dollar B2B contract

KAUST — Coral Reef Intelligence — agentic environmental digital twin, Red Sea

Byon Care — AI health-data exchange — 30% faster physician decisions

iRocker (USA) — 60% transaction growth in a 90-day engagement

7NODES — Founded, scaled to 50+ across 5 markets — acquired 2019

12 Concepts · 5 Industries

Where I'd point AI next.

Twelve build-ready concepts across the five industries I know from the inside — a third proven in production, the rest designed for what's coming.

Shipped

Automotive · Jetour, UAE

Agentic Showroom OS

Why —: The retail journey restarts at every touchpoint — context dies between sales, test drive, and aftersales.
What —: A multi-agent copilot spanning sales, test drive, aftersales, and loyalty.
How —: Central reasoning layer over inventory, DMS, and CRM; voice + text; human handoff with full context preserved.
Impact —: Continuity across the entire ownership lifecycle — the journey never restarts.

Future-State

Automotive

Fleet Predictive Maintenance at the Edge

Why —: Fixed-interval servicing wastes money and still misses failures.
What —: On-vehicle edge agents predicting degradation from live telemetry.
How —: Quantised models on edge hardware; agents schedule optimal maintenance windows against fleet utilisation.
Impact —: 20–30% fewer unplanned breakdowns; proactive operations replace reactive scrambles.

Future-State

Automotive

Dynamic Pricing & Inventory Intelligence

Why —: Dealer pricing runs on instinct and lags the market.
What —: An ML pricing engine over demand signals, competitor moves, and inventory age.
How —: Agents adjust listings within operator guardrails; every change measured for elasticity.
Impact —: 5–8% revenue lift and faster inventory turns.

Shipped

Entertainment · Disney Parks

Experience Personalisation Engine

Why —: Rich guest data, generic experience — conversion lags what the data makes possible.
What —: Real-time personalisation across journey, offers, and checkout.
How —: A rigorous experimentation framework; only validated changes roll out.
Impact —: 40% conversion lift — sustained by a framework that keeps compounding.

Future-State

Entertainment

Agentic Content Discovery

Why —: Static recommendations miss mood, context, and taste evolution.
What —: A conversational discovery copilot that learns taste through dialogue.
How —: Preference agents build evolving taste profiles; serendipity is tuned, not random.
Impact —: Engagement and retention lift; a premium discovery tier opens new revenue.

Future-State

Entertainment

Venue Digital Twin & Guest Intelligence

Why —: Parks and giga-projects fly blind on real-time guest flow.
What —: A live digital twin of the venue with agentic crowd, queue, and energy orchestration.
How —: Sensor fusion + simulation + agents that rebalance staffing, F&B, and attractions in real time.
Impact —: Higher per-guest spend, shorter queues, lower operating cost.

Shipped

Healthcare · Mitchell International

LLM Claims Intelligence

Why —: Manual first-pass review throttles the entire adjudication pipeline.
What —: Autonomous triage, anomaly flagging, and queue prioritisation.
How —: Confidence thresholds + human-in-the-loop escalation, evaluated against adjuster benchmarks.
Impact —: 30% faster processing with the first manual pass eliminated.

Future-State

Healthcare · Builds on Byon Care

Predictive Risk & Care Orchestration

Why —: Health systems intervene after the crisis, not before it.
What —: ML risk models over EHR + social determinants, with agents routing high-risk cohorts to care coordinators.
How —: Pre-analysed patient context delivered inside the clinician workflow.
Impact —: 20–25% readmission reduction; intervention before deterioration.

Future-State

Telecommunications · Informed by Celcom Axiata & Circles.Life

Self-Healing Network Operations

Why —: NOCs detect issues after customers feel them.
What —: Edge agents that correlate anomalies across cells and trigger resolution before impact.
How —: Autonomous load-balancing within guardrails; only novel events escalate to engineers.
Impact —: 50%+ MTTR reduction; the NOC shifts from firefighting to capacity strategy.

Future-State

Telecommunications

Precision Churn & Lifetime-Value Agents

Why —: Blanket retention offers burn margin on the wrong customers.
What —: Churn prediction with per-customer driver attribution and agent-generated retention plays.
How —: Offers optimised for LTV; high-value at-risk accounts routed to humans with full AI context.
Impact —: 5–10% churn reduction with smarter retention spend.

Future-State

Travel · Informed by Expedia & Backroads

Revenue & Pricing Intelligence

Why —: Static pricing rules lag demand shocks.
What —: ML pricing over booking velocity, events, weather, and competitor moves.
How —: Agents adjust across channels inside operator guardrails; elasticity measured continuously.
Impact —: 5–12% revenue lift with reduced markdown waste.

Future-State

Travel

Agentic Journey Concierge

Why —: High-intent travellers abandon when planning is friction-heavy.
What —: A copilot that designs, books, and live-adjusts entire itineraries from a conversation.
How —: Preference learning + booking-system integration + real-time replanning for weather, delays, discoveries.
Impact —: Browse-to-book conversion and bundled AOV lift; a concierge tier opens premium revenue.

Foundations

Education & recognition.

Education

BSc (Hons) Information Technology with Business Information Systems
Middlesex University, United Kingdom
2007 – 2010
BE Computer Science
Birla Institute of Technology International Centre, Bahrain
2006 – 2007

Recognition

Leaders in Innovation Award · 2017
Insight Success · Top 15 Companies in India
Top 30 Influential Business Leaders · 2019
Insight Success · as CEO of 7NODES
Featured in Entrepreneur Magazine · 2016
World Startup Expo — one of Asia's largest startup events

Beyond the Role

Where I invest my thinking.

Digital Experience

Products people love to use — AI that feels effortless, never heavy-handed. The craft of making complex systems feel simple.

AI & Agentic Systems

Multi-agent architectures and frontier models, moved past demos into value. The discipline of production over promise.

Startups & Venture

Founder with an exit. $25M portfolio managed, 20+ startups shaped. Open to advising teams solving genuinely hard problems.

Advisory & Mentorship

Helping leadership teams unstick stalled AI initiatives and build durable innovation muscle inside the organisation.

Get in Touch

Let's build what's next.

If your AI initiative needs to move from promising to production — that's the conversation I want to have.

nebu@meetnebu.com LinkedIn

Dubai, UAE · Serving the Middle East, South East Asia & India · Remote and on-site.