Cognitive Substrate for Modern Organizations

The layer underneath your AI tools.

95% of AI projects fail — because no foundation knows what your company has decided, learned, and built. Engram is that foundation: an AI-native operating layer that turns institutional knowledge into the finished work each person needs. Permission-filtered. Customer-hosted. Your model. Outputs flow back; intelligence compounds.

Read the architecture
Multi-tenant from day one Federated by design Self-hosted-first Apache-2.0 Community Edition forever

Your AI investment is evaporating — not because the models fail, but because nothing remembers across the conversations.

$0.0T
Annual loss to information overload, global enterprise
McKinsey Global Institute
0.0
Hours per knowledge worker per week, spent searching for information
McKinsey + IDC
$30-60K
Friction tax per knowledge worker per year
Atlassian, Asana, HBR aggregated
0%
Of enterprise AI projects fail to reach production scale
MIT NANDA Project + Gartner 2026
What Engram saves you

Five scenarios where APAC decision-makers lose time + money today.

96% of APAC enterprises will boost AI spend by 15% in 2026. 83% of APAC CFOs say AI is reshaping finance. Yet 95% of AI pilots fail to deliver ROI — because the substrate underneath is missing. Here is what changes when the substrate is there.

CEO · CFO

Sales numbers now — from every source.

Tuesday morning. CEO needs Q3 pipeline by segment, last week delta. Data lives across Salesforce, HubSpot, Excel, Slack channels, customer-success tool. Today: 3 days of RevOps work. Board moment gone.

With Engram: Multi-source substrate-query in seconds. Salesforce + HubSpot + Slack-conversations aggregated, with citations. CEO answers in the meeting itself.

Today 3 days
With Engram 30 seconds
CTO · Head of People

A senior leaves — the knowledge stays.

When a senior engineer leaves: 48% team-efficiency drop short-term. 42% of their expertise is undocumented and permanently lost. Replacement runs at 52% efficiency for 6 months. APAC SME multi-firm holdings hit hardest — smaller teams, higher concentration.

With Engram: Architecture decisions, reasoning, disagreements live in the substrate. When the senior leaves, the institutional memory stays. New hires onboard via substrate access, not Slack archaeology.

Today 6.2 months
With Engram Weeks

Dogfooded: SynaptiQ has run as NGIS's internal cognitive substrate for 18 months — the proving ground for everything Engram inherits.

CISO

Eliminate Shadow AI — with a sanctioned alternative that is actually better.

81% of workers use unsanctioned AI tools. Singapore PDPA penalty: SGD 1M or 10% turnover (whichever higher). PDPC late-2025 precedent: "Negligence during digital migration is no longer a valid defense." Malaysia PDPA 2024: mandatory DPO + 72h breach notification.

With Engram: The sanctioned alternative that is genuinely better than shadow AI — fast (5-min install), joyful UX, customer-hosted, audit-logged. CISO unlocks budget independent of productivity-ROI debate.

Risk today SGD 1M penalty
With Engram Audit-logged
CFO · CHRO · Compliance Lead

PDPA + MAS + BNM compliance — native in the architecture.

Singapore PDPA + IMDA Model AI Governance + Agentic AI Framework + MAS FEAT/Veritas + Malaysia PDPA 2024 + BNM RMiT. Six weeks of audit prep per cycle. Compliance officer + outside counsel + IT — six-figure cost annually.

With Engram: Audit artefacts are byproducts of substrate operation, not reconstructions. Article-level mapping native (PDPA Section 24, MAS FEAT, BNM RMiT). Right-to-be-Forgotten as first-class lifecycle-op, not 6-month engineering project.

Today 6 weeks
With Engram Hours

NGIS Pte Ltd is Singapore-based — PDPA + MAS + IMDA from first-hand experience. The Scigate BEC forensic-IR case (2026) was preserved at evidentiary grade on the SynaptiQ substrate.

Why Engram

Six patterns that drain every organization, everywhere.

The Engram team studied the cognitive friction tax across eight industries and four regions. The same six failure modes appear in every one. They are not productivity problems — they are substrate problems. Engram solves them at the layer where they live.

01 / Decision Amnesia

The room agreed. The system forgot.

Decisions made in conversation never reach the artefacts that downstream teams act on. New hires re-litigate questions the company already answered three years ago.

02 / Evidence Drift

The number was right. The source was lost.

By the time a claim reaches the board deck, no one can find the calculation, the assumption, or the analyst who built it. Confidence collapses on the first hard question.

03 / Departmental Silos

Eight tools. Eight memories. Zero coherence.

Sales has Salesforce. Engineering has Jira. Compliance has Confluence. Each holds a fragment. Nothing holds the whole. The knowledge worker becomes the integration layer.

04 / Disagreement Loss

The minority view didn't survive the meeting notes.

Dissent gets sanded off in the summary. A year later, when the consensus turns out wrong, no one remembers who said so — or why they were dismissed.

05 / Compliance Audit Tax

Six weeks to reconstruct what should be a query.

PDPA. MAS audit. BNM RMiT. SOC 2. EU AI Act-ready. Each demands provenance Engram already has natively. Bolt-on compliance costs six figures per audit cycle — or worse, the audit fails.

06 / Context Window Collapse

Your LLM forgets faster than your interns.

Every new prompt is amnesia. Multi-turn reasoning collapses past 128K tokens. The substrate that gives LLMs durable memory is the missing piece — not better models.

Multi-Audience Native

One substrate. Four scales. No bolt-ons.

Most cognitive tools serve one audience — the individual, the team, or the enterprise. Engram is the first substrate designed multi-tenant-multi-user from Layer Zero. The same knowledge graph reaches from a single department all the way to peer federation between independent organizations.

Scale 01 / Department

50–500 users. One organizational lens.

The smallest unit. Sales. Engineering. Legal. Each gets its own knowledge graph, its own retention policy, and its own audit log — while sharing the substrate underneath. Decisions made here surface upward when (and only when) they should.

SME multi-firm holdings start here. They often run four to ten departments in parallel from day one.

Capacity
50–500

active users
one organizational lens

Topology
What you get

Sales sees Sales.

Every claim, decision, and dissent scoped to the department. No cross-leak. No accidental knowledge bleed when someone moves teams.

Why you can rely on it

7-year retention. Audit isolated.

SOC 2 + PDPA evidence built-in. Right-to-erasure scoped per department. Retention policy configurable per team.

Scale 02 / Branch

Multiple departments. One regional roof.

A regional office. A factory site. A national operating unit. Cross-departmental memory at branch scope — without sharing it back to HQ unless the data-residency rules say so. The substrate handles that distinction natively.

For multinational SMEs and Fortune-class regional HQs, this is where the federation logic earns its keep. Singapore data stays in Singapore. Frankfurt data stays in Frankfurt.

Scope
Multi-dept

data-residency aware
regional sovereign

Topology
What you get

Singapore asks Singapore.

Cross-departmental memory at branch scope. Singapore data stays in Singapore. Kuala Lumpur data stays in KL. The substrate enforces the boundary — not the application.

Why you can rely on it

Data-residency cryptographic, not policy.

Zero cross-region leaks since deployment. Auditor-queryable boundary. APAC regulatory zones honoured natively (MAS Singapore, BNM Malaysia, HKMA Hong Kong, APRA Australia).

Scale 03 / Subsidiary

Multiple branches. One legal entity.

Where compliance lives. Where audit happens. Where governance is enforced. A subsidiary is the unit your regulators ask about — and the unit your CFO consolidates. Engram makes both queries native, not bolt-on.

Fortune-class industrials typically begin here. They standardize Engram at the subsidiary layer first, then federate up.

Governance
One entity

where audit happens
regulators ask

Topology
What you get

Where the regulator asks.

The unit your CFO consolidates. The unit your auditor queries. Branches roll up under one legal entity — one append-only log, one access boundary, one governance policy.

Why you can rely on it

PDPA. MAS audit. SOC 2.

Native. Not bolt-on. PDPA Singapore + Malaysia, MAS audit packages, SOC 2 Type II. GDPR + EU AI Act ready for global expansion. Six-figure audit cycles collapse to a query.

Scale 04 / Peer-Federation

Independent organizations. Shared substrate.

The substrate-level innovation. Two enterprises — competitors, partners, supply-chain peers — can federate selectively without either ceding control of its data. A pharma sponsor and its CROs. An OEM and its tier-1 suppliers. A consortium of banks under shared regulatory pressure.

This is what GitLab is to git, what ServiceNow is to workflow: an open substrate that becomes the connective tissue between organizations that otherwise cannot speak.

Pattern
N peers

sovereign · selective
cryptographically scoped

Topology
What you get

Pharma shares with CROs.

Banks share under consortium rules. Suppliers share evidence — without sharing pricing. The substrate is the connective tissue between organizations that otherwise cannot speak.

Why you can rely on it

Cryptographically scoped. No PII leaked, ever.

Selective claim publishing. Peer-by-peer policy. Trust keys per partner. Every shared claim auditable from both sides — nothing happens off the record.

NGIS Track Record

Engram is the team's fifth production system.

The top of a stack we have built layer by layer over four years. MeshWave wires the field data. Cognify turns data into operational intelligence. SynaptiQ has run as our internal cognitive substrate for 18 months. Engram is that substrate, productised. Four shipped product lines, production-deployed with Fortune 500 and institutional clients across SEA.

Product 01
Connectivity & Capture

MeshWave

Days to deploy. Years to outlive.

Industrial LoRa-Mesh IIoT — wireless production-floor communication and mission-critical sensor networks, multi-region and multi-tenant from day one. Long-range, self-organising peer mesh: drop nodes in place and the network forms itself — no gateway, no IT integration, no cabling. Rollout in days, not weeks. Battery-powered with multi-week autonomy. One stack drives status indicators, Andon terminals, and sensor nodes.

  • StatusProduction
  • TopologyPeer mesh
  • PowerBattery, multi-week
  • StackMulti-tenant Layer 0
Product 02
Connectivity & Capture

AeroScale

Real-time weigh-in. Audit-grade trail.

Wireless aircraft weight-and-balance for heavy maintenance and overhaul — in the hangar for months, or out on the apron. Ultra-low-power load cells run 90 days per charge, streaming over MeshWave LoRa. Technicians pick aircraft type and model from a complete database; the software visualises live and produces the audit-ready report in seconds — no paper-to-Excel detour. No fixed scales. No cabling. No paper ticket. Customer-hosted, same peer-mesh stack.

  • StatusProduction
  • DevicePortable
  • NetworkMeshWave LoRa peer
  • Audit90-day data trail
Product 03
Analytics & Intelligence

Cognify

The data ISO forces. The insights they actually need.

ASEAN manufacturers already log every cycle, reject, and stoppage — ISO and IATF compliance demand it. The data exists. The BI engineer does not. Cognify reads what they already capture and turns it into a structured intelligence report: findings for management, prioritised actions for the shop floor — written in 6 minutes, in the tone the reader actually needs. Six analysis areas, from OEE to custom.

  • StatusPilot Q3 2026
  • Areas6 analysis areas
  • OutputFindings + Actions
  • AudienceOperator → C-Level

NGIS Track Record is presented as evidence of team capability and design discipline. MeshWave, AeroScale, Cognify, and SynaptiQ are distinct NGIS product lines — not Engram customers, not Engram traction. Engram is pre-launch.

Pricing

Transparent. Per-seat. Customer-hosted.

No Engram Cloud — ever. Software license, you host. Apache-2.0 Community Edition forever free. Paid tiers add federation, governance, SLA, and the things Fortune-class enterprises actually need. Annual prepay.

Community
Single tenant. OSS.
$ 0 Apache-2.0
forever
  • Self-hosted, customer infra
  • Full feature parity (memory model)
  • Community Discord support
  • No seat limit
Starter
10–49 seats min.
$ 60 per user
per month
  • Multi-tenant ready
  • Email + Slack support
  • Single subsidiary scope
  • SSO included
Profile‑2 (mid)
500–2,500 seats. Fortune-class.
$ 45 per user
per month
  • Full 4-scale architecture
  • Federation primitives included
  • SLA 99.95% + on-call
  • PDPA + MAS + BNM + EU AI Act-ready
  • Dedicated Forward Deployed Engineer
  • Quarterly business review
Profile‑2 (top)
2,500+ seats. Multi-subsidiary.
$ 40 per user
per month
  • Peer-federation included
  • SLA 99.99% + 24/7
  • Sovereign data residency
  • Dedicated FDE + Account Executive
  • Executive sponsor + roadmap input
  • Custom security audits
  • White-glove migration

Annual prepay. Customer-hosted only — no Engram Cloud. Bring your own open-weight LLM. Apache-2.0 Community Edition stays free, forever. Transparent volume pricing — no "Contact Sales" wall.

About NGIS

An AI-native company. Two years of substrate discipline. Three production deployments.

NGIS PTE LTD is a Singapore-incorporated industrial-technology company building the Cognitive Substrate for modern organizations — the infrastructure layer that lets enterprises remember, reason, and audit their AI decisions across teams, tools, and time.

We are AI-native, not AI-first. The distinction matters: AI-first companies bolt AI onto existing products; AI-native companies build the architecture around AI from Layer Zero. Multi-tenant from day one. Federated by design. Compliance-by-construction. Customer-hosted only.

Founded in 2024 and operating from Singapore with production deployments across Malaysia, NGIS already ships three software systems running with Fortune 500 and institutional clients — MeshWave (LoRa-Mesh IIoT), Cognify (AI manufacturing intelligence), AeroScale (aircraft weighing for Malaysia Airlines at KLIA). Engram is the substrate layer we have been building underneath them.

3
Production deployments with Fortune 500 / institutional clients in Malaysia
SGD 300K
Seed closed April 2026 — Nextier Ventures, Singapore
4
Shipping product lines — MeshWave, Cognify, AeroScale, SynaptiQ (internal)
2 yr
From incorporation to production-deployed substrate — ASEAN-rooted
FAQ

The questions buyers actually ask.

Is Engram cloud or on-prem?
Customer-hosted only. There is no Engram Cloud — not now, not later. You run it on your infrastructure, your cluster, your sovereign region. We sell software licenses, not SaaS subscriptions. The Apache-2.0 Community Edition is free forever and feature-complete for the memory model. Paid tiers add federation, governance, and SLA.
Which LLM does Engram use?
Whichever you bring — and most enterprises mix. Engram itself stays customer-hosted; the LLM is your choice per team and per use case. For compliance-sensitive workloads, IBM Granite leads on-prem in 2026 — first open-weight family with ISO 42001 certification, Apache 2.0, with IBM IP indemnification and Granite Guardian for bias and hallucination detection. Llama, Mistral, Qwen, and Gemma also run on your own infra. For non-sensitive quality-first tasks like graphics, code, or marketing — Anthropic Claude (ISO 42001 + SOC 2 + HIPAA-BAA), Google Gemini, OpenAI. We help you map model to workload, with PDPA, MAS Veritas, AI Verify, and EU AI Act in mind.
How is this different from Mem0, Letta, Cognee, or Zep?
Those are agent-memory primitives — library-level, single-tenant, developer-first. Engram is substrate-class — multi-tenant-multi-user from Layer Zero, federated by design, compliance-by-construction. Different category. We respect the work in the agent-memory layer; we are building the layer underneath it. Cognee at €7.5M seed, Mem0 at $24.5M cumulative, Letta at $10M @ $70M post-money are the closest comparables. None are substrate-class. None have entered Series B.
Why federated? Most organizations aren't asking for that.
Today, no. In 18–36 months, yes. We are building for the moment when a pharma sponsor needs its CROs to share evidence without ceding data, when a consortium of banks needs shared regulatory memory without merging, when supply-chain peers need to coordinate without exposing competitive surface. Federation is the substrate-class moat. Build it now or rebuild later.
What does "Compliance-by-Construction" mean?
PDPA Singapore + Malaysia, MAS audit, BNM RMiT, IMDA Model AI Governance, SOC 2 — with EU AI Act, GDPR, and HIPAA ready for global expansion. The audit artefacts these regimes demand are byproducts of Engram's normal operation, not bolt-on instrumentation. The substrate's append-only event log, source attribution, and dialectical provenance are the same primitives an auditor asks for. You do not pay six figures per audit cycle to reconstruct what Engram already has.
When does Engram launch?
Engram's pilot program is open now and commercial GA is H1 2027. We are taking 5–10 founding customers as early adopters — direct founder access, weekly working cadence, founding pricing locked, and real influence on what ships at GA. Two ways to engage below: join the waitlist for product updates and pilot announcements, or apply directly as a founding customer.
Will Engram use our data to train models?
No. Customer-hosted only — your data never leaves your infrastructure, never reaches us, never trains anything else. The LLM you plug in is yours; you can swap it without re-papering. Data ownership stays 100% with you, by architecture, not by promise.
What happens to our deployment if NGIS folds?
Apache-2.0 Community Edition is the full memory model — free forever. If NGIS stops shipping tomorrow, your deployment keeps running on the existing license. We sell governance, federation, SLA, and support on top of OSS. The right to remain functional is yours by license, not by vendor goodwill.
Does Engram integrate with our existing stack?
Yes — the integration list is extensible by design. Native connectors include Slack, Confluence, Jira, Salesforce, SharePoint, Workday, Outlook, ServiceNow, GitHub. Anything else in your stack plugs in via webhook or REST API. SSO from Starter tier; SCIM and RBAC from Profile-1. Engram lives inside your existing stack — feeds data in, returns finished deliverables back to the tools your team already uses.
Be among the first

Two ways to engage. Both real.

Engram pilot program runs now through GA in H1 2027. Join the waitlist for updates and pilot announcements, or apply directly as a founding customer for paid pilot access.

Quick

Join the waitlist

Be the first to know when Engram is ready. Pilot announcements, founder briefings, product updates. Unsubscribe anytime.

Serious
5–10 founding customer slots

Apply as founding customer

Direct founder access. Weekly working cadence. Founding pricing locked. Real influence on what ships at GA. Paid pilot from day one.