Edge AI · MI-EAI-06
Secure Contextual Generative AI Workstation
An isolated node pulls context from MQTT/Modbus and the RAG store to draft shift reports and RCAs, with RBAC, audit log and zero outbound traffic.
Edge AI · MI-EAI-06
An air-gapped node that pulls live context from MQTT and Modbus and from the plant's document store to draft shift reports and root-cause summaries — with role-based access, an audit log and zero outbound traffic.
- ₹1,71,000pilot BOM · hardware only
- 8 – 12 daysto a live pilot
- Advancedintegration level
- 3validated providers
Why GenAI in a regulated plant is a governance problem
The problem
Teams want generative AI to draft shift reports, work instructions and root-cause summaries from live machine data and plant documents, inside a regulated or air-gapped environment. The value is obvious; the governance — who asked what, what data it saw, what it produced — is the blocker.
- GenAI drafting needs live plant context, which cannot leave site
- No audit trail for prompts, data used or outputs
- Role-based access to sensitive data sets is unsolved
- Regulated environments require an isolated, verifiable node
The solution
An isolated node behind a hardware firewall on its own VLAN, with disk encryption and signed boot, runs the local LLM and RAG stack (MI-EAI-04/05) plus context connectors that read MQTT topics and Modbus registers. A governance layer enforces role-based access to document sets and data sources, logs every prompt, the context retrieved and the output, and exposes templates for shift reports, work instructions and RCA drafts.
- Live context from MQTT/Modbus plus the document store
- Templates for shift reports, work instructions and RCA drafts
- Role-based access to data sets; prompt, context and output audit log
- Hardware-isolated, encrypted, no outbound route
Validated service providers
Validated by TwoElectrons and ranked by validation score: verified credentials, completed jobs on the platform and client ratings.
- Sample Provider S — Bengaluru · On-prem LLM & RAG · ★ 4.9 · 12 jobs
- Sample Provider T — Hyderabad · TinyML & edge inference · ★ 4.7 · 9 jobs
- Sample Provider U — Pune · Secure AI infrastructure · ★ 4.6 · 7 jobs
Provider listings are samples pending live registrations.
Pilot BOM and cost
Reference hardware to run the pilot. Unit costs are indicative Indian market prices for the component class.
| Qty | Item | Unit cost | Line |
|---|---|---|---|
| 1× | Edge AI compute box GPU/NPU class, as MI-EAI-04 | ₹1,40,000 | ₹1,40,000 |
| 1× | Hardware firewall / air-gap network switch VLAN isolation, no WAN route | ₹25,000 | ₹25,000 |
| 1× | TPM / hardware security key disk encryption, signed boot | ₹6,000 | ₹6,000 |
| 1× | Local LLM + RAG stack as MI-EAI-04 / 05 | Free tier / open source | — |
| 1× | Context connectors & governance layer MQTT/Modbus adapters, RBAC, prompt & output audit log | Free tier / open source | — |
| Pilot BOM total | ₹1,71,000 |
POC BOM cost only — hardware for a pilot. Installation, integration and provider services are quoted separately. Request for Quote
How a secure contextual GenAI node works
- Isolate. Deploy the node on its own VLAN behind the hardware firewall.
- Connect. Point context connectors at MQTT topics, Modbus registers and the document store.
- Govern. Configure roles, data-set access and the audit log.
- Draft. Generate shift reports and an RCA draft; review with the plant lead.
Pilot architecture
Live context (MQTT · Modbus · documents) → Secure GenAI node (LLM + RAG + governance) → Air-gap boundary (VLAN · firewall · TPM) → Reviewed drafts (shift report · RCA · WI)
The node reads plant context and produces drafts inside the boundary; every prompt, context retrieval and output is logged.
Business case: report time, audit trail, isolation
Ranges are typical figures reported for this class of solution; your pilot establishes the numbers for your plant.
- Minutes to a shift report or RCA draft (from live data and documents)
- 100% of prompts, context and outputs logged (auditable by role)
- 0 outbound routes (VLAN, firewall and TPM verified)
- 8 – 12 days to a working node (one plant)
Who this is for: Plant heads and operations excellence, Information security and compliance, Pharma, chemicals, defence and aerospace suppliers, Power and critical infrastructure, Regulated multi-site groups, Any site that needs GenAI with an audit trail.
Illustrative case study
Illustrative scenario · not a client reference
A defence-component supplier with an air-gapped shop floor in Bengaluru (illustrative)
Set-up. Secure node on a dedicated VLAN behind a hardware firewall, TPM-backed disk encryption, connectors to the MES MQTT feed and the OEM manual store, roles for production, quality and management.
What happened. Shift reports that had taken supervisors 30–40 minutes were drafted from live data in under two minutes and reviewed in five. A quality escape RCA was drafted from the deviation record and the relevant SOPs in one session, with every source cited in the audit log for the customer's review.
- Pilot budget: ₹1,71,000 hardware
- Shift report time: 35 → 7 min
- Audit coverage: 100% of sessions
FAQ
How is 'air-gapped' verified?
The node sits on a VLAN with no route to the internet behind a hardware firewall; the firewall logs are reviewed with IT during the pilot and the configuration is documented.
Can it write to the MES or PLC?
No — connectors are read-only by design. Outputs are drafts for human review.
Who can see what?
Role-based access controls which document sets and data sources each role's prompts can retrieve; the audit log records every retrieval.
What does the ₹1,71,000 cover?
The compute box, firewall/switch and security hardware; the software stack is open source. Governance configuration, hardening and provider services are quoted after a Request for Quote.
