RUDX Pitch Deck

Edge AI that guards the assets nobody is watching.

Unmanned infrastructure security · Bangalore, India

rudx.ai 01

The problem

India is covering its landscape with unmanned, powered, theft-exposed assets.

EV chargers, fuel stations, telecom towers, solar farms, substations. The usual defence is a human guard who witnesses theft instead of preventing it, at ten times the cost of monitoring it with software.

$1.3–1.9B

Estimated India cable & battery theft losses per year

₹45–75K

Monthly cost of one 24×7 guard post (three shifts)

18%

India public EV chargers non-functional, partly theft-driven

02

Why now

Four tailwinds arrived at once.

Theft wave

Copper near record prices. Charger cord-cutting up 2.7× year over year; every new site is a new target.

Asset boom

EV charging must grow ~45× by 2030. Millions of new unmanned, powered sites.

STQC / BIS

From Apr 2026, uncertified Chinese CCTV is restricted. A compliant, India-built stack becomes a procurement requirement.

DPDP Act

From 2027, on-device processing is a legal advantage. Edge-first is privacy-by-design.

03

The solution

An edge-AI box on the cameras already there.

RUDX understands live video on-site — no cloud streaming — and does what a guard can't:

Understands behaviour in plain language: "someone tampering with the cable bay."
Fuses video with asset telemetry — charger status, SCADA — so "cable cut + person + no active session" is a near-certain alert.
Deters in real time: floodlight, siren, voice, before the cut.
Proves it: FIR- and insurance-ready evidence packages.

The shift

From record-the-theft to prevent-the-theft.

Cloud video analytics records what happened hours later. RUDX acts on-site in the seconds that matter, on hardware that costs less than one month of a guard.

04

The technology

Frontier video understanding, shrunk to a $60 chip.

Our edge is a family of ultra-compact models we distil from frontier AI down to run continuously on commodity NPUs under 10 watts — open-vocabulary video intelligence where it has never fit before.

9,000×

Model compression demonstrated end-to-end, kept 91% of the teacher's decision quality

0.23 ms

Inference latency on our production edge NPU

<10 W

Always-on power draw — fanless, survives Indian field conditions

<$10K

Total training compute for the full model program

05

Where we are

A prototype running in the real world today.

Working prototype deployed on-site, EV charging as the first vertical.
Hardware proven on the target board — detection, deterrence, and recording, end to end.
Pipeline validated: the distillation and edge-deployment stack is measured, not theoretical.
Deployment-grade models on the roadmap for H1 2027, gated by staged go/no-go milestones.

The deterrence loop

Camera → on-device understanding → telemetry fusion → light · siren · voice → evidence packet. All on the box. No site needs bandwidth for cloud video; only events and 10-second clips leave over 4G.

06

Market · India, bottom-up

One engine, many verticals.

EV charging (2030)₹2,000+ Cr

Telecom towers₹1,600 Cr

Warehouses₹1,000+ Cr

Fuel retail₹840 Cr

Solar + C&I₹500+ Cr

₹6–9K Cr

TAM — all unmanned powered assets, ~2030

₹3–4K Cr

SAM — five verticals we sell into in 5 years

₹100–140 Cr

SOM — year-5 ARR, 40–60K sites at ~₹2,000/mo

Site count × documented ARPU · directional 07

Business model

Recurring revenue, hardware bundled in.

Tier 1 · Self-serve

₹500–1,000 /site/mo

Edge alerts to owner app, clip storage, zones. No human desk.

Tier 2 · Fusion + triggers

₹1,500–3,000 /site/mo

Telemetry fusion, automated deterrence, evidence packages. The core product.

Tier 3 · Human-in-loop

₹3,000–6,000 /site/mo

Verified alerts, escalation SLAs. Added only when ARR pays for the desk.

70%+

Gross margin at scale — edge caps cloud & bandwidth spend

$150–200

Edge box cost, bundled as HaaS and recovered inside the contract

>3:1

Target LTV:CAC via channel distribution, not direct sales

08

Competition

Nobody owns telemetry-fused theft prevention for these assets.

Player	What they do	Gap we exploit
Securens (IN)	Guard-replacement e-surveillance, 22,000 bank sites	Banking-locked, capex-heavy, no EV/energy/telecom, no fusion
Hikvision / CP Plus	₹3,000 NVRs with free motion analytics	Detect only — no verification, deterrence, or evidence; STQC-blocked
Spot AI / Hakimo (US)	AI on existing cameras, talk-down, $62M / $20.5M raised	Not in India, no telemetry fusion, not edge-efficient enough here
RUDX	Edge AI + telemetry fusion + deterrence + evidence	The India-conditions, edge-native, fusion-first stack — proven abroad at $100M+ ARR

09

What compounds

The box isn't the moat. Five things are.

01 · Fusion

OCPP / SCADA / NOC integrations built once, reused forever. Each one raises switching cost.

02 · India models

Dust, monsoon, sodium light, stray animals — every labelled false alarm is training data rivals must repurchase.

03 · Evidence

FIR packets, insurer formats, chain-of-custody. Process integrations are slow to copy.

04 · Trust

Deterrence lives or dies on the siren. Reliability reputation is the brand in a reference-driven category.

05 · Flywheel

More sites → more edge cases → better precision → lower ops cost → cheaper tiers → more sites.

The structural bet

ARR businesses aren't killed by cheaper boxes. They're killed by churn — and fusion + evidence kill churn.

10

Go to market · channel-first

Sequenced verticals, distribution not door-to-door.

Y0–Y1 · WEDGE

EV charging

Prove three numbers — theft reduction, false-positive rate, working evidence pack. Then embed in the charger BOM.

Y1–Y2 · NEXT

Fuel retail

Inherit OMC relationships. One standardization = tens of thousands of sites.

Y2–Y3 · SCALE

Towers + solar

Single accounts worth 100,000+ sites. SCADA fusion is the same engine, new adapter.

Y3+ · EXPAND

Warehouses

Higher ARPU, rapid-deploy units, once the core engine is proven.

The whole game is CAC: one OEM integration puts RUDX inside every unit shipped; one tower-co deal is 100,000+ sites. Direct mid-market is where Indian video-AI goes to grind — we don't play there.

11

Team

Built by an operator who ships hard systems.

II

Indrajith Indraprastham — Founder

Close to a decade building AI-driven platforms and large-scale distributed systems, as engineer and founder.

Gojek · clickstream analytics, $6.5M/yr saved FAANG · agentic-AI latency optimization Mayo Clinic · HPC / SLURM in HIPAA CTO · CiaraAI Poetry (34k★) contributor ML publication · genre classification in/indrajith-indraprastham ↗

A lean technical team, staged milestones with explicit kill criteria, and capital committed only behind evidence.

12

The ask · Shell E4

Shell sits inside our two lead verticals.

1

Pilot access

Instrument Shell fuel-retail and Shell Recharge sites for theft reduction, false-positive rate, and evidence quality — our proof and your loss-prevention data.

2

Sector distribution

Introductions to OMC retail-engineering and CPO teams to shorten a slow enterprise sales cycle.

3

Mentorship & capital readiness

Energy-procurement guidance and support toward a 2027 seed raise.

Indrajith Indraprastham · indrajith@rudx.ai · rudx.ai · LinkedIn

13