Platform
Drones and sensors collect. Dimlang is building what comes after: a platform that ingests mission data, organises it by place and time, applies analysis, and puts evidence in front of the people who decide.
01Architecture
Each stage hands a more structured product to the next. The final stage belongs to people: the platform supports operational response, it does not perform it.
Drone, EO, IR, sensors, mission files
OutputRaw mission data
Secure transfer and normalisation
OutputNormalised records
Mission · location · time · sensor
OutputIndexed mission store
Geospatial analysis and computer vision
OutputCandidate observations
Analyst workspace
OutputReviewed evidence
Analyst feedback loop. Confirmations and rejections return to analysis as evaluation data.
Human operational response
OutputDecisions by accountable people
02Operational data foundation
AI is only useful when the data underneath it is sound. Dimlang is not a chatbot over video files. It is a mission-data platform first, designed so that every output can be checked against what was actually collected.
{
"id": "obs_01J8Q4M2",
"mission": "M-014",
"sortie": 2,
"source_asset": "ir/M-014/s2/frame_018442.tiff",
"captured_at": "06:42:18+01:00",
"sensor": "thermal",
"geometry": { "type": "Point", … },
"footprint": { "type": "Polygon", … },
"detector": "person-candidate",
"model_version": "research-0.3",
"confidence": 0.71,
"review": { "state": "pending", "by": null },
"provenance.sha256": "9f1c…e27a"
}Original frames, video, and telemetry are kept exactly as collected. Every derived product points back to them.
Clocks, coordinate systems, and sensor formats differ between platforms. They are reconciled once, at ingestion.
Every frame and observation is indexed by where it was seen, so the map is the primary way in.
Observations are placed on a single timeline across sorties and missions, so repeat passes can be compared.
Each observation records its source asset, sensor, model version, and processing steps.
Access, review decisions, and exports are logged, so the record can be examined after the fact.
Encrypted at rest and in transit, inside the boundary the deploying institution defines.
Confirmations and rejections are captured as data, for evaluating and improving models over time.
A model output is never shown without the frame it came from. No answer without its source.
03Analyst experience
Analysts work where the observations are: on the ground they cover, across the missions that saw them, with the source frame one step away.
04AI and analytics
Models sit inside the pipeline, between organised mission data and human review. Their job is to narrow the search, not to deliver verdicts.
“AI should reduce the amount of data an analyst must inspect — not obscure where an answer came from.”
Models rank and surface. They do not decide. Their job is to shrink the haystack.
Every candidate opens to the original frame, its position, time, and sensor.
Nothing leaves the workspace as a finding until a person has reviewed it.
05Deployment
Mission data from sensitive operations often cannot leave a facility, a network, or a jurisdiction. Dimlang is architected to be deployed inside those boundaries rather than asking data to cross them.
Managed infrastructure where policy allows it.
Dedicated tenancy under the institution's control.
Inside the institution's own facilities and network.
Near the sensor, where bandwidth is scarce.
Intermittent or no connectivity, synchronising when possible.
Every environment is reproducible from versioned definitions.
The same services run in a cloud region or a server room.
No component requires exposure to the public internet.
Integrates with the institution's identity provider; access follows role and need.
Data encrypted in transit and at rest, with keys held by the operator.
Who saw what, when, and what they decided.
Health, throughput, and processing lag are measurable in every deployment.
Data stays in the jurisdiction and facilities the institution chooses.
No hard dependency on a single cloud provider.
These are design goals of an architecture in development. Dimlang does not currently hold security certifications or accreditations; these would be pursued against the requirements of each deploying institution.
06Capability status
Dimlang is an early-stage organisation. Nothing on this site describes a fielded deployment, a customer, or an operational result.
07Responsible use
For government, research, technology, public-safety, and search-and-rescue organisations.