ZetaIQ — Telemetry, Visualization & Analytics for Robot Fleets

The problem

Right now, robot issues are invisible

Three habits keep them that way: the data sits in silos, the analysis starts after the fact, and triage is done by hand.

Fragmented data in silos

Logs on the robot, metrics in one tool, recordings on a disk, video somewhere else. Nothing shares a clock, so nothing can be correlated.

Reactive analysis, telemetry pulled by hand

An issue is noticed hours later. Then someone SSHes into the robot, copies logs and bags off it, and reconstructs what happened from whatever survived. The sensor context is usually already gone.

Manual issue triage

Every flag is read by a person, compared against yesterday from memory, and traced to a cause by trial and error. A firmware version quietly degrading a third of the fleet looks like unrelated faults until someone joins the dots.

What goes in

Every signal a robot makes, on one timeline

Text logs, numeric series, spans, topic recordings and camera video are different shapes of data. ZetaIQ Telemetry stores them as one multimodal record of the fleet, time-aligned from the first byte.

Logs

Robot and service logs, tagged by fleet, robot, site and build. Search across every site at once, or drop from a trace span straight into the lines it emitted.

Traces

End-to-end traces across robot software and the cloud services it calls. See where a mission spent its time, and which hop was the slow one.

Metrics

Counters, gauges and histograms from robots and backends. Dashboards and custom charts over any series, at fleet or robot scope.

Recordings

rosbag and MCAP topic recordings and camera video, captured continuously, on a trigger or around an event. Replayed on the same timeline as everything else.

Start with the standard OpenTelemetry collector

Logs, traces and metrics ship through the open-source OpenTelemetry collector: no proprietary agent, no re-instrumentation. The ZetaIQ collector adds per-robot rate limiting, store-and-forward and recording capture when you want them.

Tenant and fleet identity are stamped server-side from the device certificate, never sent by the client.

Works with what you run

Built on the standards your robots already speak

Open collectors in, open views out. Nothing about your stack has to change to start, and nothing locks you in later.

ROS 2 / ROS 1

Topics, rosbag and MCAP from any distro

Nodes log and emit metrics through the OpenTelemetry collector on the robot; recordings are captured and uploaded alongside.

OpenTelemetry

OTLP logs, traces and metrics

The standard collector and SDKs in C++, Python, Rust, Go and more. If it speaks OTLP, it is already instrumented for ZetaIQ.

.mcap · .bag

Recording formats

Upload, index and replay recordings on the fleet timeline. Continuous, triggered or contextual capture around an event.

Grafana

Bring your own dashboards

Point your existing Grafana at stable SQL views over your data. Same data, same meter, no second pipeline.

Your tools

Use your own visualization tools

Open any indexed recording, or any view over your data, in the tools your team already uses, from the exact moment on the timeline.

SELECT … FROM v_metrics

SQL and BI tools

Any tool that speaks SQL: notebooks, BI dashboards, spreadsheets. Stable views are the public schema, so queries keep working.

Slack · PagerDuty

Alerts where your team is

Route alerts to Slack, PagerDuty, webhooks and email by fleet, severity or on-call rotation, with the context attached.

export · API

Your data stays yours

Query it, export it, or stream derived metrics back into your own systems. Usage is visible per fleet, in the same platform.

ROS, OpenTelemetry, MCAP, Grafana, Slack and PagerDuty are trademarks of their respective owners, named to describe compatibility.

How it fits together

A collector on the robot. One platform behind it. Any front end.

Robots / edge

OpenTelemetry collector

or the ZetaIQ collector

logs · traces · metrics
rosbag / MCAP capture
per-robot rate limiting, store-and-forward

on each robot, or on an edge box at the site

ZetaIQ Telemetry

Ingest gateway device identity · quotas · rate limits · read / write audits · fleet-level access control

Time-series store

logs · traces · metrics · derived metrics

Object store

recordings · video · reports

One tenant · one clock your data, isolated per customer, queryable and exportable

Consumed by

ZetaIQ visualization

timeline replay · dashboards · drill-down

ZetaIQ analytics

health flags · root cause · reports · Ask

Bring your own

Grafana · BI tools · SQL over the same views

You run one collector. Ingest, storage, analytics, dashboards, backups and scaling are managed for you, so your engineers stay on your product.

What you can do with it

Replay the whole fleet. Plot anything against it. Ask it questions.

Scrubbing a site timeline in 2D and 3D.

Any instant, whole fleet

Scrub to a moment and see every robot, handoff point and zone at once, in 2D and 3D, with occupancy and work rate per area over any window. A factory line, a hospital corridor, a farm block or a fulfilment hub: the same replay.

Plotting a service's latency against fleet throughput.

Any signal vs. performance

A robot's battery, a service's latency or a firmware rollout, plotted against throughput and waiting time on one axis. Fleet state at an instant, not a daily average.

Asking the fleet a question in plain language.

Ask in plain language

Ask a question; get the chart, the answer and the next question to ask. Underneath, health flags feed root-cause analysis and scheduled reports.

Cross-signal search: one click from symptom to cause

Every pivot is time-aligned, so you never reconstruct context by hand.

alert→log line→trace→recording frame

Analytics and automatic issue triage

A base model of how a fleet flows, fitted to yours.

ZetaIQ does not wait for a dashboard to be read. A continuous model of the operation flags deviations, ranks causes and writes the report.

Base domain knowledge ZetaIQ

How robot work flows anywhere: missions, handoff points, zones and queues; arrival, dwell, utilisation, starvation, and what healthy looks like for each.

Your dependencies Yours

Which services, robots, people and machines depend on which, scouted from raw telemetry by model-assisted inference, with confirmation where a signal is ambiguous.

Model patches Together

Where your operation differs from the base: metric recipes, attribution rules, thresholds and new flows. Versioned with your tenant.

Insights model

Reprojectsignals into entities, queues and flows

Detectdeviation against the healthy baseline

Correlategroup findings by causal chain

Rankroot cause separated from consequence

deterministic and continuous — no LLM in the hot loop

Issue triage

A health flag becomes a ranked root-cause chain: which robots, which service, which build, cause separated from consequence, on one timeline from symptom to span to log line.

Bottleneck reports

Where the fleet lost its minutes: robots vs. people vs. the equipment they hand work to (a weld cell, a ward elevator, a dock door, a charging bay), distance to each one's ceiling, causes ranked by impact. Narrated, per site, on a schedule.

The base is generic and domain-blind, so it works on day one. The dependencies and patches are what make the findings yours, and a patch on your tenant never changes anyone else's model.

Built for every kind of robot

If it moves, senses or serves, it belongs here.

The model underneath knows queues, missions and dependencies, not one kind of machine. Any robot that emits telemetry, and every service it talks to, lands in the same record.

AMRs and AGVs

Moving goods, parts and samples across a site.

Autonomous forklifts

Pallet handling, racking, trailer loading.

Robot arms and cells

Welding, assembly, machine tending, picking.

ASRS and shuttles

Grid storage, shuttles, lifts and ports.

Drones and UAVs

Inspection, surveying, crop and stock counts.

Humanoids and legged robots

General-purpose work and patrol.

Delivery and field robots

Sidewalks, farms, mines, campuses.

The services behind them

Fleet managers, planners, APIs, cloud backends.

Industries: warehouse and logistics · manufacturing · agriculture · healthcare · retail · defense · research labs

How it is priced

Pay for data, not for robots.

Two meters, both visible inside the platform. Add robots, sites and users without touching the bill.

Telemetry — per GB

Ingest and storage

Every signal type: logs, traces, metrics and recordings. One meter, whether you query through ZetaIQ or your own Grafana.

Analytics & ETL — per compute unit

Insights, reports, derived metrics

The insights model, scheduled reports and your metric recipes run on compute units sized to your fleet.

SML

No per-robot fee
No per-seat fee
Bring-your-own Grafana included
Data export included

Usage is telemetry too. You see the same numbers we bill against, per fleet, inside the platform.

Exact pricing is worked out against your data volume and compute needs.

Who we are

Built by the team that ran the fleets.

The team has deployed thousands of robots in production and scaled deployments from zero to thousands of robots. This is the platform we wished we had while doing it.

18+ yearsbuilding and operating robot fleets

Thousandsof robots deployed in production

0 → thousandsdeployments scaled from the first robot to fleet scale

10+patents in fleet management and robot coordination

Fleet management at scale

The team built the fleet-management core shared by autonomous forklifts and AMRs: distributed multi-robot route planning, traffic control, and coordination with people and other equipment on the same site.

End-to-end robot stacks

Architects of complete AMR and ASRS stacks, from perception and navigation to optimal task allocation, taken from first prototype to thousands of robots in production.

Observability, first-hand

Set up the telemetry infrastructure and dashboards for 30+ services, with a pipeline handling millions of events per minute from thousands of robots. First responders on live fleets, and the authors of the root-cause analyses that went back to customers.

ZetaIQ Robotics Private Limited · founded 2026 · building the telemetry, visualization and analytics platform for robot fleets.

hello@zetaiq.io LinkedIn

Questions

Frequently asked

Does ZetaIQ Telemetry work with ROS 2 and ROS 1?

Yes. Topic recordings arrive as rosbag or MCAP files, captured continuously, on a trigger, or around an event, and are indexed and replayed on the same timeline as your logs, traces and metrics. Logs and metrics from ROS nodes flow through the standard OpenTelemetry collector.

Do I need to install a proprietary agent on my robots?

No. Start with the standard open-source OpenTelemetry collector and the OTLP exporters you may already use. The optional ZetaIQ collector adds per-robot rate limiting, store-and-forward and recording capture, and you can switch to it later.

Can I keep Grafana and the dashboards I already have?

Yes. Point Grafana, or any tool that speaks SQL, at the stable views over your data. Bring-your-own visualization is included and reads the same data at the same cost as the ZetaIQ apps.

Which recording and video formats are supported?

rosbag and MCAP for topic recordings, plus camera video segments. Recordings are stored in your tenant's object store, indexed by robot, mission and time, and replayed on the fleet timeline or opened in your own visualization tools.

How is my data isolated from other customers?

Every customer gets their own database and object storage behind a hard boundary, not a configuration flag. Tenant and fleet identity are stamped server-side from the device certificate, so a client can never write into another tenant. Reads and writes are audited.

Where does it run? Can it run in my own cloud?

ZetaIQ Telemetry runs as a managed service today. Private-cloud deployment in your own VPC and air-gapped deployment are on the roadmap. Your data is exportable at any time.

How is ZetaIQ Telemetry priced?

Telemetry is priced per gigabyte ingested and stored; analytics and ETL are priced per compute unit. There is no per-robot and no per-seat fee, and bring-your-own Grafana is included. Exact pricing is worked out against your data volume and compute needs.

How do I get started?

Two ways: book a 30-minute walkthrough of a simulated site with nothing to install, or point one staging fleet's telemetry at us and get the first replay and report on your own data. Use the early-access form and we will schedule an intro call.

Limited early access — 2026

Get early access

We are onboarding a small group of early customers. Two ways to start:

A

See it first

A 30-minute walkthrough of a simulated site and the product. Nothing to install.

B

See it on your data

Point one staging fleet's telemetry at us through the standard OpenTelemetry collector. Get the first replay and report on your own fleet.

See what your robots can't tell you.