PILLAR 03 · AI PRODUCTS & PLATFORMS
AI Products & Services
Enterprises are sitting on more data and AI ambition than they can operationalize: video data goes largely unused, and institutional knowledge sits scattered across systems, drifting further from reality with every release cycle. Forar Tech through their partnership with Parabola9 closes that gap through two independent, proprietary platforms, SPECTRA for video intelligence and SENTRA for enterprise intelligence, backed by full-stack
SPECTRA
A Gen AI video intelligence platform delivering near real-time incident detection, natural language Q&A on footage, and severity mapped alerts across hundreds of cameras.
- Near real-time detection: across 100s of camera feeds simultaneously
- Natural language search: ask footage a question, get an answer
- Severity mapped alerting: routes the right incident to the right team
- Enterprise correlation: cross-camera de-duplication reduces false positives at scale
- Deployed across manufacturing, retail, transportation,
healthcare, and smart cities
100s
CAMERAS / DEPLOYMENT
5
VERTICALS LIVE
PLATFORM 02 · ENTERPRISE INTELLIGENCE
SENTRA
Smart Enterprise Neural Transformation & Reasoning Agents: an agentic AI platform combining advanced RAG, knowledge graphs, a conversational interface, and persistent enterprise memory.
- Advanced RAG engine: grounded in governed enterprise sources
- Persistent memory: compounds knowledge over time, not per session
- Autonomous agents: act on data, not just answer questions
- Governed access: aligns with enterprise security and compliance requirements
- Built for knowledge management, document analysis, compliance, and decision support
99%
DOC PROCESSING ACCURACY
1,000+
USERS, ANCHOR DEPLOYMENT
01
Proprietary IP, not integrations
SPECTRA and SENTRA are ai-built, patent-pending platforms, not repackaged open source tooling stitched together.
02
Agentic, not just generative
03
Proven, quantifiable outcomes
60%+ reduction in review cycle time, 95%+ false positive reduction, and 99% document processing accuracy in production.
04
Full stack AI engineering
From custom model development to MLOps, the platforms are backed by the services needed to deploy and run them at scale.
END-TO-END AI SERVICES
BUILD & MODEL
- Custom model development
- Conversational AI & small language models
- RAG & vector databases
- Intelligent agents & content summarization
- BI & visualization
- MLOps / AI engineering
WHY FORAR TECH
AI is one part of an end-to-end manufacturing value chain spanning Engineering (CAD/CAE/PLM) through IT (ERP, IoT, data warehousing) to AI: capturing shopfloor signal, warehousing it, and applying AI across machinery, shopfloor, and business layers for a complete design-to-decision cycle. A talent-first delivery model keeps attrition well below India IT-industry norms, and the majority of Forar Tech’s clients are referenceable.
SPECTRA — A Closer Look
SPECTRA is an Agentic Vision Fabric: an intelligence layer that fuses video with external signals (voice, weather, crime, demographics, IoT and operations data) to connect the dots, predict risk and automate preventive action. A unified surveillance platform built on advanced Large Language Models and proprietary, patent-pending technology, it delivers near real-time incident detection, human-like understanding of scenes, actionable insights and conversational interaction with video footage.
WHY IT’S TIME TO RETHINK SURVEILLANCE
1.5B+
HOURS OF CCTV FOOTAGE GENERATED DAILY
98%
OF THAT FOOTAGE IS NEVER REVIEWED
95%
OF INCIDENTS MISSED AFTER 22 MINUTES OF MANUAL MONITORING
100s
OF CAMERAS PER DEPLOYMENT, ANY EXISTING INFRASTRUCTURE
Most AI surveillance systems still rely on motion detection and rule-based triggers that lack contextual awareness — leaving the safety gaps that matter most.
KEY CAPABILITIES
01
Object detection & tracking
Detects and tracks people, vehicles and objects in real time to sharpen situational awareness.
02
Contextual intelligence
Understands scene context — who, what, where and why — for accurate threat interpretation and event summarization.
03
Connected intelligence
Links siloed cameras and systems into a unified network, enabling shared intelligence and multi-camera event correlation.
04
Automated agentic workflows
Automates repetitive surveillance tasks, freeing human operators for critical decisions.
05
Conversational interface
Just ask, and SPECTRA answers —
transforming how teams interact with surveillance footage.
06
Unified platform for video insights
Centralizes all video data into a single dashboard with near real-time insights, analytics and access across devices.
WHY CHOOSE SPECTRA
Human-like understanding
Analyzes footage with a level of understanding that closely mimics human cognition.
AI-powered incident investigation
Natural-language queries and instant, clear
summaries cut investigation time significantly.
Camera agnostic
Deploys over any and all existing camera
infrastructure.
Scalable & API-ready
Supports 100s of cameras with modular, easy
integration.
Map-based alerts
Visual alerts, prioritized by severity, highlighted on a site map.
Flexible deployment
Runs on the cloud or on-prem, whichever the site requires.
Industry Use Cases
Manufacturing
- Ensure worker safety by monitoring compliance with safety protocols.
- Monitor Hazardous areas to restrict unauthorized entry/access.
- Near real-time incident detection and emergency Response and Evacuation
Traffic & Smart Cities
- Near real-time detection of traffic violations and accidents.
- Optimize urban traffic flow with analytics-driven insights.
- Enhance public safety by monitoring high-risk zones.
Senior Living
- Detect falls and ensure immediate
assistance for residents. - Monitor compliance with safety
protocols in shared spaces. - Proactively monitor and raise safety concerns to prevent incidents .
Security & Surveillance
- Identify unauthorized access and potential threats proactively.
- Summarize hours of footage into key actionable events.
- Ensure compliance with safety and security guidelines.
Retail
- Identify unauthorized access and potential threats proactively.
- Summarize hours of footage into key actionable events.
- Ensure compliance with safety and security guidelines.
SPECTRA for the Manufacturing Sector
Continuous monitoring of shopfloor safety, hazardous-area access, and compliance across plants and facilities; cutting incident response time and manual
monitoring overhead.
Ensure worker safety by monitoring compliance with safety protocols.
Monitor hazardous areas to restrict unauthorized entry and access.
Near real-time incident detection, emergency response, and evacuation.
Monitor
Near real-time monitoring across 100s of camera feeds simultaneously.
Detect
AI-powered detection of accidents, safety violations, and unauthorized access.
Query
Natural-language Q&A on any footage; just ask, and SPECTRA answers.
Act
Severity-mapped alerts routed to the right team for faster response.
SENTRA — A Closer Look
SENTRA is an enterprise knowledge platform that indexes, connects and activates SAP, Salesforce, Databricks and data from many other enterprise sources through AI. Built on the cloud with enterprise-grade security, it delivers organizational intelligence without sacrificing speed or accuracy: every SAP record, Salesforce order and database instantly connected and queryable.
THE QUESTIONS SENTRA ANSWERS
Your enterprise applications have the answers. One platform, multiple data sources; all queryable in seconds.
FINANCE & SPEND
Budget vs. actuals
Spend vs. approved budget by cost centre
Order-to-invoice gap
Salesforce orders not yet invoiced; revenue-leakage exposure
VENDOR & PROCUREMENT
Duplicate payments
Duplicate vendor invoices across the last 60 days
Unapproved POs
Overdue, no goods receipt — invoice exposure value
COMPLIANCE & AUDIT
Drift check
Process documents out of sync, by transport count
Auto-rewrite
Redlined PDD from every transport since Q1
CHANGE & MIGRATION
Root cause
Order errors vs. transports moved that day
Migration scope
Objects safe to retire vs. needing work
ASK IT THE WAY YOU’D ASK AN ANALYST
“If I cut OpEx 10%, which cost centres have the most suppliers, and which Salesforce deals are most at risk?”
“Which Salesforce deals closed last quarter have no invoice, and what is the total revenue leakage?”
This is agentic reasoning: SENTRA pulls history, status and signals simultaneously — no analyst, no waiting.
KEY CAPABILITIES
01
Ask anything, in plain English
One question spans SAP configuration,
documents and transaction history; every answer is cited with evidence at source
02
Living knowledge graph
A map of how every transaction, document, process step and owner links — built and kept current automatically from live data.
03
Governed semantic layer
Metrics such as DSO, AP aging and win rate are defined once and used by every agent; agents never touch raw field names.
04
Autonomous domain agents
Agents that review and rewrite process
documents, or clear blocked orders acrossthe flow — unattended, with an audit trail.
05
Memory that survives sessions
Facts learned on Monday are available on Tuesday; live task lists and checkpoints mean interrupted work is never lost.
06
Personas & skills per role
Tailored personas from the C-suite to the process analyst, configured without code changes.
WHY CHOOSE SENTRA
Your data, governed
A knowledge graph over SAP, Salesforce and Databricks with RBAC, PII controls and dataaccess rules built in.
Cost control by design
Five built-in mechanisms — caching,
summarisation, on-demand skills — cut runcost by 75–80%.
Agentic, not single-turn
26+ plug-and-play tools, micro-agents, persistent memory and resumable plans that act on data.
Enterprise-grade security
Azure AD / Entra SSO and RBAC, Key Vault secrets management and private virtual networks.
Live now, fast to deploy
An 8-week pilot with no enterprise licence,no data migration and no vendor lock-in.
Open, AI-ready architecture
Interchangeable LLMs and an MCP plug-in layer your EA team can review — not a black box.
Use Cases
Knowledge Management & QA
Enable employees to quickly find answers to complex questions across vast document repositories using natural language
Document Analysis
Automatically extract insights, patterns, and key information from large volumes of documents
Compliance & Risk Management
Streamline compliance processes with automated document review, policy verification, and regulatory reporting
Decision Support
Provide executives and teams with AI-powered insights and recommendations based on comprehensive data analysis
SmartHub: SENTRA for the ERP estate
A flagship ERP use case: a Semantic Layer, Knowledge Graph, Chat Agent, and autonomous PDD Agent that index and validate SAP / ERP documentation,
closing the gap between what the system does and what the documentation says it does.
01 – LAYER
Semantic Layer
A shared business meaning across raw ERP tables and process data
02 – MAP
Knowledge Graph
Connects processes, documents, and system objects into one map
03 – ASK
Chat Agent
Conversational access to process and documentation knowledge
04 – ACT
PDD Agent
Autonomously reviews and flags
documentation drift; acts, not just answers
SPECTRA Use Cases - Manufacturing
Safety Vigilance Across Every Site
Prepared for the Manufacturing Sector · September 2026
A manufacturer’s operations generate answers faster than they generate certainty about where those answers live. A maintenance backlog, a PPE violation, an unauthorized entry — each sits in a different camera feed, on a different site, under a different rule set. This document goes deeper than a summary of features: it lays out SPECTRA as a continuous, regulation-aware vigilance layer — not a checklist, but a system that watches for the precursor as closely as the incident.
PPE, OSHA/MSHA Compliance & Continuous Vigilance
Two regulatory regimes typically apply across a manufacturer’s footprint: OSHA-aligned practice in the US, with MSHA (Mine Safety and Health
Administration) wherever mining or extraction operations are involved; and ISO 45001 with local occupational health law elsewhere — together
governing processing, power generation and dispatch operations. SPECTRA’s underlying vision of AI doesn’t change between sites; only the
rule set per zone does.
WHY VIGILANCE, NOT JUST CHECKLISTS
Industry data cited in SPECTRA’s own briefing: over 1.5 billion hours of CCTV footage are generated daily, 98% of it is never reviewed, and 95% of
incidents go unnoticed after just 22 minutes of manual monitoring, due to fatigue and attention loss. A checklist audit catches what was true now someone
looked. Continuous vision AI is built for the other 21 hours and 38 minutes.
PPE detection, by zone and role
| PPE ITEM | WHERE IT’S ENFORCED | WHAT IT CATCHES |
| Hard hats / bump caps | All production, storage and yard zones | Anyone entering a marked zone without head protection. |
| Hi-visibility clothing | Haul roads, yard traffic, conveyor walkways | Reduces the leading industrial hazard: vehicle/pedestrian incidents. |
| Respiratory protection | Areas with airborne fine powder (process powders and dust) | Confirms respirator use wherever combustible or respirable dust is present. |
| Safety glasses / goggles | Chemical handling and processing units | Flags missing eye protection at fixed handling points. |
| Hearing protection | Turbine halls, compressor rooms, heavy machinery bays | Confirms compliance in defined high-noise zones. |
| Fall-arrest harnesses | Loading gantries, conveyor structures, tank tops | Detects work at height without a visible tie-off point. |
| Emergency escape equipment | Personnel entering confined or high-risk zones | Confirms that every worker carries the required emergency equipment before entry. |
Procedural and regulatory compliance
| PROCEDURE | WHERE IT APPLIES | WHAT IT CATCHES |
| Lockout-tagout (LOTO) verification | Maintenance on de-energized equipment | Confirms locks and tags are visibly applied before work starts — not just logged on paper. |
| Confined-space entry protocol | Tanks, silos, process vessels | Verifies a permit, attendant, and monitoring equipment are present at the point of entry. |
| Hot work near combustible dust | Welding or cutting near fine powder accumulation | Flags hot-work activity where visible dust buildup creates an explosion risk — a known hazard in mineral powder handling. |
| Machine guarding | Conveyors, crushers, rotating equipment | Confirms guards are in place, not removed for “quick access.” |
| Powered haulage / pedestrian separation | Haul roads, loading yards | Monitors for pedestrians in active vehicle paths — the most-cited hazard category in heavy industry. |
| Housekeeping / dust accumulation | Ledges and structural surfaces near processing units | Tracks dust buildup visually over time as an early warning before it becomes a combustible dust incident. |
Vigilance and behavior-based safety
This is the layer beyond compliance snapshots — continuous pattern watching for the precursor, not just the rule of violation.
| FOCUS AREA | WHERE IT APPLIES | WHAT IT CATCHES |
| Control-room operator alertness | Plant and utility control rooms running 24/7 | Signs of fatigue or attention loss during long monitoring shifts — the exact failure mode manual CCTV review suffers from. |
| Unsafe-act detection | Phone use near moving machinery, bypassed barriers, unsafe lifting | Catches the behaviour before it becomes the incident, not after. |
| Near-miss pattern detection | Recurring close calls at the same conveyor transfer point or crossing | Surfaces a hotspot for an engineering fix before it produces an injury. |
| Loitering / unauthorized presence | Remote and multi-site plants, off-hours | Extends effective security coverage without expanding guard headcount. |
| Tailgating at access points | Badge-controlled entries to restricted zones | Flags a second person entering one person’s credentials. |
| Lone-worker / man-down detection | Remote areas of large plant sites | Detects a worker down or motionless for an abnormal period and triggers an alert automatically. |
| Contractor & visitor compliance | Third-party crews on site for turnarounds or capital projects | Confirms visiting contractors follow the same PPE and zone rules as permanent employees. |
SENTRA Use Cases - Manufacturing
The Manufacturing Copilot
Prepared for the Manufacturing Sector · September 2026 A manufacturer’s operations generate answers faster than they generate certainty about where those answers live. A delivery status update, a certificate of expiry, a maintenance backlog — each sits in a different system or a different document. This document goes deeper than a summary of features: it lays out SENTRA as a true “ask anything” copilot across every system and document the enterprise holds.SENTRA as the Manufacturing Copilot
A single conversational interface, grounded in every connected source — SAP, Databricks, Salesforce, and every PDF, contract, certificate and report the enterprise holds. Anyone authorized asks a plain-language question and gets a cited answer back, regardless of which system or document holds it.WHY VIGILANCE, NOT JUST CHECKLISTS
Industry data cited in SPECTRA’s own briefing: over 1.5 billion hours of CCTV footage are generated daily, 98% of it is never reviewed, and 95% of
incidents go unnoticed after just 22 minutes of manual monitoring, due to fatigue and attention loss. A checklist audit catches what was true now someone
looked. Continuous vision AI is built for the other 21 hours and 38 minutes.
What the Copilot reads
The value of a copilot is set by what it can see. The answer set spans structured transactional systems and years of unstructured documents — both are in scope from day one.
SOURCE | WHAT IT HOLDS | SAMPLE QUESTION IT CAN NOW ANSWER |
SAP (Finance, MM, SD, PP) | Purchase orders, invoices, goods receipts, cost centres, production orders | “Which POs for plant equipment are open past their expected delivery date?” |
Databricks / data lake | Production telemetry, shipment tracking feeds, sensor logs, ML feature tables | “How does this week’s plant output compare to the same week last quarter?” |
Salesforce | Customer orders, opportunities, accounts, delivery | “Which customer orders placed this month haven’t been |
| commitments | scheduled for shipment yet?” |
Certificates & compliance PDFs | Certificates of Analysis, ISO and product certifications, REACH dossiers, SDS | “Pull the current Certificate of Analysis for the last batch shipped to this customer.” |
Contracts & commercial documents | Customer and vendor contracts, pricing schedules, Incoterms, renewal clauses | “Which vendor contracts renew next quarter and haven’t been reviewed?” |
Shipping & customs documents | Bills of lading, vessel manifests, certificates of origin, | “Is any shipment loading this week missing its certificate |
| import paperwork | of origin?” |
Inspection, audit & maintenance | Internal/external safety audits, insurance surveys, | “What did the last third-party safety audit at the main |
records | equipment maintenance logs | plant flag as open items?” |
Investor & governance materials | Board decks, rating-agency reports, bond disclosures | “Summarize every mention of acquisition integration |
|
| risk from the last two board packs.” |
Delivery and logistics intelligence, end to end
This is the everyday question every manufacturer asks most often, in one form or another: what is the status of this shipment, right now, and why. Today that means knowing which of several systems tracks which leg. The copilot removes that step entirely.
USE CASE | EXAMPLE QUESTION | WHY IT MATTERS |
Live shipment status, any leg | “Has today’s export shipment finished loading, and is it | Answered instantly instead of a call to the logistics |
| on schedule to depart?” | team. |
Delivery exception detection | “Which export shipments are delayed this week, and why?” | Surfaces the actual cause — customs hold, missing document, vessel delay — in one query. |
Order-to-delivery-to-invoice loop | “Which orders from last month are delivered but not yet invoiced?” | Closes the revenue-leakage gap without manual three-system reconciliation. |
Document-triggered delivery holds | “Is any shipment held up because its REACH or | Connects document intelligence directly to logistics |
| certificate of origin paperwork is missing?” | status. |
Multi-modal shipment tracking | “Where is order #8834, and which mode is it moving on — sea, rail or road?” | One question spans all three shipment modes without knowing which system tracks which. |
One copilot, every role’s own questions
The same underlying knowledge graph, scoped differently per person — each gets the questions relevant to their job, asked in their own words.
ROLE | QUESTION THEY ASK | WHAT THEY GET BACK |
Plant Manager | “What maintenance is overdue on the process pumps, and could it affect this week’s output target?” | A direct answer to linking maintenance backlog to production risk. |
Export & Logistics Coordinator | “Which shipments are loading right now, and is anything missing a certificate of origin?” | Real-time dispatch status without a phone call. |
QA / Regulatory Manager | “List every product certification expiring in the next 90 | One list spanning industrial, food-grade and regulatory |
| days, across all product grades.” | certifications. |
Procurement Lead | “Show me every open PO with a vendor whose ISO certificate has lapsed.” | Compliance risk surfaced inside a procurement workflow, not after the fact. |
HR / Safety Manager | “How many safety inspections are overdue at the plant | A live count instead of a quarterly manual audit. |
| right now?” |
|
Investor Relations | “Pull every reference to acquisition integration risk from | A cited summary in minutes, not a research afternoon. |
| the last two quarters of board and rating-agency materials.” |
|
CFO / COO | Cash, FX exposure, vendor risk and fulfilment | Covered in the companion CFO/COO executive brief — |
| performance across all entities. | this copilot layer is what sits underneath it. |
