Eight ventures. Productized.
Deployable.
Each venture is a Rubix asset: white-label-ready, methodology-grounded, scalable across environments and cultures. The methodology is the asset; the ventures are the receipts.
The methodology is the asset. Eight times over, it has produced a venture worth shipping.
01
Computer Vision · Construction
BIMLens.
Computer vision that reads construction drawings.
With Saudi Building Code overlay and bilingual annotation tolerance. Production accuracy, industrial throughput, under sovereignty.
Deploys for
73%
Drawing review cycle reduction
9–11 days → 2.5 days per package
91%
Clash detection recall
Up from 63% baseline, vs. held-out test set
22
Disciplines covered
Across 2D and BIM artifacts on giga-project scale
38K
Annotated training elements
Rubix-curated dataset, fine-tuned per project
Capability
Drawing review at production accuracy
Automated review, clash detection, and code-compliance checking on 2D and BIM artifacts. At production accuracy, industrial throughput, under sovereignty.
Differentiator
Saudi Building Code, native
SBC compliance and bilingual annotation tolerance are not bolt-ons. They are the architecture. The reason off-the-shelf vendors failed where BIMLens succeeded.
Discipline
Eval against the senior reviewer
Recall and precision benchmarked against held-out drawings reviewed by senior QA. Production threshold defined before code shipped.
Technology stack
- ·Custom-trained YOLOv8 object detection backbone
- ·Bilingual transformer layout-language model
- ·Saudi Building Code rules engine
- ·Forge API integration to Revit and Navisworks
- ·ISO/IEC 42001 governance baseline · sovereign hosting
In production
Re-work cost from missed clashes SAR 22m → SAR 6.4m annualized. Senior engineer review hours released: ~4,800 / year.
02
Enterprise AI Agent
YANA.
A single AI agent platform deployed across every function of a diversified group.
With shared retrieval, shared evals, and shared governance. Function specialists on a shared spine.
Deploys for
8
Functions on one platform
HR, Supply, Finance, Strategy, GRC, S&M, Legal, Commercial
3
Models orchestrated
Claude · OpenAI · Llama 3 sovereign
340GB
Curated enterprise corpus
Role-based access, lineage, full audit trail
80%
CoE run by client team
Hand-the-keys executed by month 12
Capability
Function specialists on a shared spine
Eight specialist agents with scoped tool access and prompt libraries. On shared retrieval, shared evals, and shared governance. One architecture, eight functions stronger.
Differentiator
Governance is not appended
NIST AI RMF and ISO/IEC 42001 mapped before architecture finalized. The risk register fills first. That is what makes board-level approval possible.
Discipline
Built to compound, not centralize
Hub-and-spoke CoE deliberately built so the client owns it by month 12. Rubix's success metric is not retention, it is hand-off velocity.
Technology stack
- ·Multi-model orchestration (Claude / OpenAI / sovereign Llama 3)
- ·Function-specific specialist agents with scoped tool access
- ·Shared retrieval over the group's curated corpus
- ·Ragas + custom function evals (legal citation, financial correctness)
- ·Governance backbone · registry, audit trail, AUP automation, IR
In production
Engineered for diversified groups running 5+ operating companies. Architecture validated, governance baseline mapped, ready for first-client deployment.
03
Edge Computer Vision
iVision.
Visual intelligence on existing CCTV.
Privacy-by-design, no facial recognition, aggregated metrics only. The architecture is the legal posture.
Deploys for
0
Personally identifiable data persisted
Frames processed on-edge, aggregated metrics only.
5
Operational layers
Footfall, queue, dwell, journey, tenant-mix.
Edge
Deployment topology
Per-asset edge inference, no cloud uplift required.
DPO-grade
Privacy architecture
Reviewed for KSA & GCC data-protection regimes.
Capability
Footfall, queue, dwell, journey
Asset-scale customer-flow intelligence without individual tracking. Engineered to be reviewed by an operator's DPO before a model touches a frame.
Differentiator
Privacy is the architecture
No facial recognition. No individual tracking. No data leaving the edge perimeter. The legal posture is what makes this deployable in the region.
Discipline
No rip-and-replace
Runs on existing CCTV. Integration via the operator's existing video management system. Net-new infrastructure cost engineered to zero.
Technology stack
- ·Edge-deployed CV pipeline on per-asset edge servers
- ·Aggregated-only metric streams · No PII persisted
- ·DPO-reviewed privacy architecture, designed for external counsel sign-off
- ·24-hour forward staffing optimization, retraining cadence weekly
- ·Tenant-intelligence dashboard for leasing teams
In production
Engineered for retail real estate operators with multi-asset portfolios. Privacy architecture reviewed and ready for first-asset deployment.
04
F&B Compliance Vision
KitchenEyes.
Real-time food-safety compliance on existing kitchen CCTV.
Eval-tuned for false-positive intolerance. The metric that matters: operators trust the alerts.
Deploys for
<7 min
Detection-to-remediation
Engineered cycle from violation to alert reaching the line manager.
5
Detection classes, fine-tuned
Hairnet, gloves, hand-washing, contamination, temperature.
12K
Held-out eval scenes
Manually labeled corpus for weekly regression and tuning.
<2.4%
False-positive target
Engineered below the threshold operators set as acceptance.
Capability
Real-time, audit-grade
Detection-to-remediation cycle engineered from 3–14 days (audit-driven) to under 7 minutes. Full evidence chain-of-custody for any regulator-facing frame.
Differentiator
Operators trust the alerts
The metric that matters: false-positive rate low enough that line staff act on alerts. Eval discipline is what produces operator trust at production scale.
Discipline
Three-tier escalation
Advisory, intervention, audit-flag. Each tier mapped to manager workflow. Not all detections are alerts; not all alerts are escalations.
Technology stack
- ·Domain-specific CV models, fine-tuned for F&B kitchens
- ·Existing CCTV integration with no new camera installation
- ·Real-time mobile alerting with 3-tier escalation
- ·Audit-grade evidence trail with regulator-ready chain-of-custody
- ·Eval harness with per-class precision/recall thresholds
In production
Engineered for F&B operators running multi-outlet portfolios. Eval harness and detection classes architected, ready for first-cohort deployment.
05
Corporate Wellbeing OS
WellB.
A Corporate Wellbeing Operating System.
Not an app, not a portal, not a benefits add-on. End-to-end measurement, diagnosis, personalization, activation, and maturity progression for the Saudi enterprise.
Deploys for
5
Pillars of wellbeing
Mental, Emotional, Physical, Financial, Purposeful.
100
Holistic assessment items
Single-session OR Knock Knock 5-a-day across 20 days.
5
Maturity stages tracked
Awareness → Engagement → Behavior → Culture → Sustainable.
AR/EN
Native bilingual experience
Knock Knock avatar built in Arabic first, with Islamic psychology layer.
Capability
Continuous operating cycle
Seven-stage loop: onboard, assess, diagnose, personalize, activate, support, reassess. Wellbeing engineered as a sustained organizational capability, not a series of disconnected initiatives.
Differentiator
Knock Knock AI avatar
The 100-question assessment delivered as five micro-questions a day across twenty days, by an AI avatar. Daily ritual, not annual chore. Native Arabic, Islamic psychology layer, Vision 2030 alignment.
Discipline
Privacy is the architecture
Saudi PDPL aligned. EAP data firewalled from employer analytics. Manager views aggregate at 5-person threshold. Employees see what they get; employers never see what they didn't earn the right to see.
Technology stack
- ·5-layer architecture: Strategic, Measurement, Activation, Engagement, Support
- ·Mobile + web admin + super-admin platforms; HRIS integrations
- ·Multi-model AI: Knock Knock avatar, predictive burnout, content engine
- ·Embedded EAP with clinical governance and Islamic counselor matching
- ·Saudi PDPL + GDPR aligned · full DPIA · AI ethics framework
In production
Engineered for Saudi enterprises and Vision 2030 entities. Five-pillar architecture, Knock Knock avatar, and PDPL-aligned governance ready for first-cohort deployment.
06
Pet-Care Super-Platform
Sulhafa.
A regional pet-care ecosystem.
Combining a clinical operating system for veterinarians, a consumer app for pet owners, and an AI bridge that turns one-off vet visits into lifetime relationships.
Deploys for
3
Pillars on one platform
Social network · clinical OS · marketplace
11
Target countries, MENA
KSA + Egypt anchor; GCC high-value expansion
3.5M+
Addressable pet owners
3,200+ addressable pet-care businesses
$1.3B
Regional pet-care market
$1.3–1.8B with double-digit growth
Capability
Three pillars, one ecosystem
Social platform connecting owners, vets, rescuers and creators. Clinical OS with full EHR/EMR for clinics. Marketplace aggregating shops, food, accessories, services. Built so each pillar feeds the others.
Differentiator
The AI bridge between vet and owner
The clinic's medical record meets the owner's daily life. Predictive care alerts, vet-blessed product recommendations, churn-risk scoring, adherence prediction. The vet becomes present in the owner's life every day.
Discipline
Mission as operating principle
Born from the founder's rescued cat, Gray. Free Rescue Hub for NGOs and rescuers. National animal-welfare data system for regulators. The platform makes commercial sense because the mission shapes the architecture.
Technology stack
- ·Mobile (iOS/Android) + web platform · multi-tenant SaaS
- ·Clinical OS: EHR/EMR, SOAP notes, billing, ZATCA, inventory, scheduling
- ·AI bridge: predictive care, recommendations, churn scoring, adherence
- ·Marketplace: storefront, transactions, 5% commission engine
- ·AWS hosting · built by Rubix Studio Jordan
In production
Featured deployment: 12-clinic group, ~28,000 pets in care. Patient retention 53% → 81%. Visit frequency 1.4 → 3.2 / year.
07
Innovation Programs Platform
SparkThon.
White-label platform for innovation cohorts.
Application clustering, mentor matching, and outcome prediction. The intelligence that turns innovation programs from theatre into compounding pipeline.
Deploys for
198
Featured cohort applicants
National entrepreneurship program, single intake
5d
Application screening cycle
Down from 3 weeks for the same applicant volume
84%
Mentor-fit satisfaction
Up from 51% in prior gut-feel matching
12 min
Evaluator prep time
Per application; from 47-min cold-read baseline
Capability
Cohorts at scale, without drowning
From application clustering through evaluator briefing through mentor matching to outcome tracking. Every additional cohort compounds the platform's intelligence layer.
Differentiator
Productized, not engagement-bespoke
SparkThon is a platform that ships, not a one-off. The intelligence layer was developed, refined, and released across multiple deployments before becoming standard.
Discipline
Intelligence in the workflow
AI is surfaced contextually inside evaluator and mentor portals. Not parked in a separate "AI dashboard" that nobody opens.
Technology stack
- ·Embedding-based application clustering by problem space
- ·LLM-generated evaluator briefs with pre-flagged due diligence
- ·Algorithmic mentor matching on capability gap and sector experience
- ·Cohort outcome prediction at 6- and 12-month gates
- ·Productized evaluator and mentor portals
In production
Application screening 3 weeks → 5 working days. Mentor-fit satisfaction 51% → 84%.
08
Consulting Operating System
Rashad.
How Rubix runs its own practice.
Methodology-embedded agents, institutional memory, quality-gate automation. Before Rubix sells AI transformation, Rubix built one for itself.
Deploys for
Lifecycle
Engagement coverage
From opportunity assessment through delivery to governance.
Eval-gated
Release discipline
Every agent regression-tested against held-out historical engagements.
890
Prior engagements indexed
Institutional memory, role-based access enforced.
Internal
Deployment scope
Rubix engine, applied to Rubix work first.
Capability
The full consulting lifecycle
Methodology-embedded agents from opportunity assessment through proposal to delivery to governance. Engineered so every Rubix engagement runs through the same operating system.
Differentiator
We dogfood the methodology
Before Rubix sells AI transformation, Rubix builds one for itself. The eval discipline we recommend is the one we apply to our own consultants.
Discipline
Provenance on every output
Every Rashad-assisted deliverable carries provenance metadata. For client transparency and for our own knowledge governance.
Technology stack
- ·Methodology-embedded workflow agents per consulting phase
- ·RAG over Rubix's full engagement archive (890 prior)
- ·Quality-gate automation against firm standards
- ·Eval-driven release discipline · Regression vs. historical engagements
- ·Audit trail of consultant + AI work, role-based access
In production
Engineered for the consulting lifecycle. Methodology library, agent architecture, and eval harness designed against the firm's own engagement archive.
Every venture walks the same path.
Eight ventures. Eight cases. Same methodology applied eight different ways. When you engage Rubix, the work follows the same discipline.