Your data never leaves your perimeter.
For engineering teams who can't send proprietary documents to third-party AI vendors. SpecSense ships in three deployment tiers so you can run the exact same engineering-review intelligence inside the perimeter that your clients, regulators, and policies require.
Why deployment choice matters
Engineering documents — RFQs, P&IDs, HAZOPs, vendor datasheets, specifications — contain commercial, operational, and sometimes safety-critical information that must not leave the company perimeter. Shipping them to a hosted AI service means losing control over:
- Where the data physically sits
- Who at the vendor can access it
- Whether it's used to train future models
- What happens when the vendor is breached, acquired, or shuts down
SpecSense was built for this constraint. Three deployment tiers let you run the exact same engineering-review intelligence inside your own perimeter.
Three deployment tiers
Tier 1
SwiftModa SaaS
Fastest to value. SpecSense running in SwiftModa's EU/Frankfurt tenant with managed LLM APIs.
- Where it runs
- SwiftModa EU/Frankfurt (Supabase + Vercel)
- Setup time
- 15 minutes
- Data residency
- EU (Frankfurt), AES-256 at rest, TLS 1.2+ in transit
- AI access
- OpenAI + Anthropic + Google via paid API tiers (no training on submitted data)
From
£299/month
Tier 2
Private Cloud
SpecSense containers deployed into a VPC inside your AWS, Azure, or GCP subscription.
- Where it runs
- Your cloud tenant (AWS / Azure / GCP, any region)
- Setup time
- 1–2 weeks
- Data residency
- Wherever you provision — EU, UK, UAE, Saudi, US all supported
- AI access
- Azure OpenAI Service, AWS Bedrock, or Google Vertex AI — in your tenant, zero egress to third-party vendors
From
£1,500/month + your cloud compute
Tier 3
On-Premise / Air-Gapped
Physical server(s) in your datacentre. No inbound or outbound internet required.
- Where it runs
- Your rack, your building, your country
- Setup time
- 3–6 weeks
- Data residency
- Air-gapped — your perimeter, no exceptions
- AI access
- Local open-weight models (Llama 3.3, Mistral Large, Qwen 2.5 72B) on 1× H100 or 2× A100
From
£35,000 install + £800/month support
Controls that apply across every tier
Regardless of where SpecSense runs, these commitments hold.
No cross-tenant data flow
Each customer's data is isolated at the database layer. Postgres row-level security on SaaS; separate instances on Private Cloud and On-Prem.
No training on your data
SwiftModa does not use customer content to train models. LLM providers we interface with explicitly exclude paid-API data from training per their commercial terms.
Full audit logs
Every review request, model call, output, and user action logged with timestamp and user ID.
CEng sign-off available
Reviewed-and-signed tier delivers outputs stamped by a UK Chartered Engineer (CEng MIMechE).
DPA on signing
Standard Data Processing Agreement available at contract signing. Custom clauses negotiable at Private Cloud and On-Prem tiers.
Export + delete, anytime
Full data export on request. Full erasure within 30 days of written request. No lock-in.
Which tier fits your environment?
Answer each question honestly. The tier with the most "✅" matches is your starting point.
| Question | SaaS | Private Cloud | On-Prem |
|---|---|---|---|
| Already running cloud workloads in your own tenant (AWS/Azure/GCP)? | — | ✅ | — |
| Client contract forbids data leaving a specific region or tenant? | ❌ | ✅ | ✅ |
| Corporate 'no third-party AI vendors' policy? | ❌ | ⚠️ | ✅ |
| Environment air-gapped or behind government-grade firewall? | ❌ | ⚠️ | ✅ |
| Can you give 1–2 weeks for setup? | — | ✅ | ✅ (3–6 weeks) |
| Want lowest ongoing cost? | ✅ | ⚠️ | ⚠️ |
| Want zero infrastructure management on your side? | ✅ | ❌ | ❌ |
Default recommendation: start on SaaS for a 30-day evaluation on non-sensitive documents. If output quality earns a wider rollout and your security team requires it, migrate to Private Cloud. On-Prem is for environments where perimeter constraints are hard requirements.
Procurement questions, answered
Where exactly does our data sit?
SaaS: Supabase EU-Frankfurt (AWS eu-central-1). Private Cloud: wherever you provision. On-Prem: your rack.
What encryption?
AES-256 at rest, TLS 1.2+ in transit, all tiers. Keys managed by Supabase on SaaS; you hold them on Private Cloud and On-Prem.
Who at SwiftModa can see our data?
SaaS: a small named engineering group with time-limited, logged, break-glass access only for support escalations. Private Cloud: zero direct access — we troubleshoot via support ticket and your screen-share. On-Prem: zero direct access, ever.
What if SwiftModa gets acquired or shuts down?
SaaS: standard SaaS provisions apply (export + migration). Private Cloud and On-Prem: you own the deployed binaries. Install packages are escrowed — the system keeps running without us.
Is this SOC 2 / ISO 27001 certified?
Not yet — SwiftModa Ltd is a 2024 UK company (SC808510) and formal certification is on the 2027 roadmap (see /security). For regulated-industry buyers, this is where Private Cloud or On-Prem compensates: you inherit SOC 2 controls from your own cloud or datacentre rather than relying on ours.
Can we see the source code?
Private Cloud: inspection access via screen share during install. On-Prem: full source handover included in the install engagement.
Walk through your environment
20 minutes with a Chartered Engineer. We'll map SpecSense against your perimeter, data-residency constraints, and procurement checklist.
SwiftModa Ltd (Company No. SC808510) · Registered in Scotland · 15 Persley Den Pl, Woodside, Aberdeen, AB21 9GR, UK