01
On your server
LOCAL VISION
A fast on-server pass profiles each image on ordinary hardware — and is pluggable to a local object detector — so the expensive reasoning only runs on what matters.
- Fast
- Low-cost
- Structured output
00 Field intelligence / Kenya
POINT.CAPTURE.DECIDE.
Turn a photograph of any location into a structured business decision—what is there, what it means, who is competing nearby, and what to do next.
01 A camera that thinks
LOCBOT / RETAIL / MVP
FROM “I’M LOOKING AT A PLACE” TO “HERE’S WHAT I SHOULD DECIDE.”
Location assessment in Kenya still depends on sending someone out to walk, observe, take notes, and write a report. LocBot turns the camera already in every agent’s hand into a consistent intelligence layer.
SHOPFRONT0.96
SIGNAGE0.94
OPEN ACCESS0.87
OPTICAL READ What is physically here
- 01Retail shopfront96%
- 02Visible mobile-money signage94%
- 03Open customer access87%
- 04Stocked interior78%
- 05Dedicated parkingND
The vision read reports what is physically present. Interpreting what it means for the decision is a separate, deliberate step.
02 How LocBot reads a place
VISION → CONTEXT → DECISION
CHEAP EYES.
EXPENSIVE THINKING.
USED WHERE IT COUNTS.
A deliberate division of labour keeps every capture fast and economical. The local model observes. Geographic data supplies context. Claude reasons.
−1.286436.8172
GPS / REVERSE GEOCODING / NEARBY FEATURES / COUNTY CONTEXT
02
Reasoning layer
CLAUDE REASONING
The detections, original image, and geographic evidence become a viability score, competitor read, operational forecast, and clear recommendation.
- Interprets
- Explains
- Recommends
03 The decision
INTELLIGENCE CARD / 00:18
RETAIL VIABILITY
ILLUSTRATIVE ANALYSIS82/100
Strong roadside visibility and active commercial movement. Validate pedestrian flow before committing to fit-out.
- Foot traffic
- High
- Competitor pressure
- Moderate
- Road visibility
- Strong
- Parking
- Limited
- Estimated fit-out
- KES 420K–680K
RECOMMENDATION / 01
Proceed to a seven-day footfall validation. Position the entrance toward the main pedestrian approach and avoid direct price competition with established mini-markets.
PHOTO ✓GEO ✓KNBS BASELINE ✓CLAUDE ✓
04 Built for the field
OFFLINE FIRST / ANDROID FIRST
NO SIGNAL IS
NOT A DEAD END.
Photos, GPS, timestamp, compass heading, and notes queue locally. When connectivity returns, every capture resumes its route to an Intelligence Card.
- 01CAPTUREPhoto, GPS, heading, time.
- 02QUEUEEncrypted locally if offline.
- 03ANALYZEVision, geo, and reasoning.
- 04RECEIVERealtime completion signal.
- 05SHAREPDF or secure link.
05 One pipeline. Many field decisions.
13 SECTORS · LIVE
SAME MACHINERY.
NEW QUESTION.
Storefront viability, nearby competition, access, visible demand, and rough setup economics.
06 Built to be sold, not just demonstrated
ORGANISATION SCOPED
FIELD SPEED.
ENTERPRISE
BOUNDARIES.
For the agent, it is one capture button. For the organisation, it is scoped data, traceable evidence, role controls, and a repeatable decision process.
- 01
- Organisation isolationA bank’s records remain structurally separate from every other customer.
- 02
- Role-aware accessField agent, analyst, and administrator responsibilities remain distinct.
- 03
- Evidence provenanceCapture, location, census baseline, and AI reasoning remain visible.
- 04
- Export and sharingGenerate a clean report or a time-limited secure link.
DJANGO REST FRAMEWORKPOSTGISCELERY + REDISCHANNELSFLUTTER
07 Start from what the field can see
EVERY PLACE
HAS A SIGNAL.
Give your field teams the ability to turn what they see into what your organisation should do next.