Geo engine with AI components

Geocoding AI

Enrich GPS coordinates from photos, Excel tables or CSV files with addresses automatically. Run-based, with AI-generated report and configurable file naming.

Features

Upload

Upload your own files or manage a customer data room with personalized upload links.

Processing mode & configuration

Pick reverse, forward or normalization; the matching configuration appears inline.

Run

Start processing and follow progress live.

Reports & admin

Review reports, browse run history, delete with audit proof, all in one place.

Upload

Two ways to feed data into a run: upload locally yourself, or issue a customer a personalized upload link. Both paths land in the same processing stream.

Choose inbound path

Local upload

Upload your own files (photo ZIP, Excel/CSV with lat/lon, address list). Directly into the run.

Customer data room

Issue a personalized upload link to external customers. Magic link with expiration date, audit trail, auto-unpack.

Files

Drop files here or click to select

Drop GPS sources (photo ZIP, photo set, or Excel/CSV with lat/lon columns). Max 500 MB per file.

Archives are unpacked automatically. Supported archive formats: .zip .tar.gz .tar.bz2 .7z .rar (via libarchive). Contents are passed into the run; no manual preparation needed.

Secure storage for customer run data. Personalized upload links with expiration date, automatic token invalidation when the dataset is deleted, complete audit trail. Data lands directly in the S3-compatible object storage and is selectable from the pre-run configuration.

Active links3(today) File sets3(30 days) Pending1(operator) Revoked1(30 days)

New upload link

Create a personalized link with expiration date and token.

Active links 4

Issued upload tokens, valid and unused.

File sets 3

Incoming customer uploads, ready for run configuration.

Audit history 12

Full token lifecycle including GDPR references.

Customer / project	Created	Valid until	Status	Upload link	Files	Actions

Customer / project & files	Count	Size	Upload	Actions

Token JTI	Customer / project	Action	Timestamp	Actor	GDPR reference

Processing mode & configuration

What do the modes mean? ↗

Pick a mode; the matching configuration appears inline below (just like the upload-path switch Local / Customer data room). Pattern builder, output format and mode-specific options are pre-filled per mode.

Pick mode

Reverse geocoding

Geo to address. GPS coordinates (EXIF photo, Excel/CSV with lat/lon) are resolved to a postal address including house number.

Forward geocoding

Address to geo. Address lists are placed as lat/lon coordinates, with a confidence score per hit.

Address normalization

Raw address strings are normalized to an industry-common quasi-standard (house number, suffix, postal code, city). Default profile: FTTH.

Configuration · Reverse geocoding

Engine region

Input coordinate system (CRS)

Standard for EXIF photos is WGS84. Surveying data is usually ETRS89 / UTM.

Expected input format

Excel/CSV with columns lat, lon (also latitude / longitude) · GeoJSON file · or photo set / ZIP with GPS EXIF.

Address normalization standard (output)

Output format

File naming pattern (output)

Click a placeholder or drag and drop it into the pattern.

{original} {street} {house_no} {house_no_suffix} {postal_code} {city} {district} {state} {kgs8} {country} {lat} {lon} {run_id} {timestamp} {ext}

Live preview

Input: IMG_4521.jpg
Result: IMG_4521.jpg

Configuration · Forward geocoding

Engine region

Expected input format

Excel/CSV with columns street, house_no, postal_code, city, or alternatively a combined address column.

Output coordinate system (CRS)

Address normalization before geocoding

Recommended on, lifts the hit rate significantly.

Output format

File naming pattern (output)

In forward mode the geo tokens {lat} / {lon} are available as the result.

{original} {lat} {lon} {street} {house_no} {house_no_suffix} {postal_code} {city} {kgs8} {country} {run_id} {timestamp} {ext}

Live preview

Input: Koenigstr. 12, 70173 Stuttgart
Result: Address.xlsx

Configuration · Address normalization

Engine region

Address normalization standard

The FTTH profile splits house number and suffix and delivers KGS8.

Output format

Duplicate detection

Expected input format

Excel/CSV with a free address column or single fields (street, postal_code, city).

File naming pattern (output)

In normalization mode all normalized fields are available as tokens. {house_no_suffix} is newly separate.

{original} {street} {house_no} {house_no_suffix} {postal_code} {city} {district_part} {district} {state} {kgs8} {country} {run_id} {timestamp} {ext}

Live preview

Input: koenigstrasse 12a, 70173 stgt
Result: Koenigstrasse_12_a_70173_Stuttgart.xlsx

Run

Run summary

Status

Ready

Waiting for start

Items

200

estimated from upload

Token / cost estimate

31.05 EUR

200 initial + 7 retry passes

ETA

~2:14 min

based on 142 ms / item

Ready to start

200 items prepared · region DE · mode: Geo to address · output: enrichment + report

0 % ETA: not started yet

--:--:-- Waiting for run start...

Reports & admin

Review the current run, search the run history, delete data GDPR-compliantly and fetch audit proofs, all in one place.

Pick a section

Current run

Detail report for the latest or selected run, including KPIs, item table and download.

Run history 5

All previous runs with status, date and operator, filterable by time range.

Delete

Auto-cleanup policy or remove single runs on demand. Audit proof is generated automatically.

Audit proofs 4

GDPR-compliant deletion PDFs with SHA-256 and operator signature, available for re-download.

Geocoding AI · Run report

Field photos Q2-2026

Run run-20260511-1430 · 2026-05-11 14:30 · executed by Stephan Klahr

200

Items

193

Successful

96.5 %

Hit rate

2:14 min

Runtime

Summary (AI-generated)

The run processed 200 field photos with GPS coordinates from April 1 to June 30, 2026. 193 photos (96.5 %) could be mapped to a unique address. Three clear clusters show up geographically: the Greater Stuttgart area (84 photos, mainly the inner-city postal codes 70173 to 70182), Hamburg North (47 photos, mostly Eppendorf and Winterhude), and the Munich southern area with 36 photos. Seven photos returned no unique hit: five due to inaccurate coordinates in rural areas, two due to corrupt EXIF data. Reprocessing these seven cases via image comparison is recommended.

KPIs

200

Items processed

193

Successful

Reprocessing recommended

94.3 %

Avg confidence

142 ms

Avg latency per item

2:14 min

Total runtime

Distribution by federal state

Item details (excerpt)

Item	GPS	Result	Confidence	Status
IMG_4521.jpg	48.7758, 9.1829	Stuttgart, Koenigstr. 12, 70173	97 %	OK
IMG_4522.jpg	48.7762, 9.1834	Stuttgart, Koenigstr. 14, 70173	96 %	OK
IMG_4530.jpg	53.5763, 9.9926	Hamburg, Eppendorfer Weg 88, 20259	94 %	OK
IMG_4534.jpg	48.1373, 11.5754	Munich, Marienplatz, 80331	91 %	OK
IMG_4587.jpg	50.1186, 8.6802	Frankfurt, Zeil 42, 60313	68 %	low confidence
IMG_4612.jpg	0.0000, 0.0000	(no match)	-	EXIF error
IMG_4613.jpg	52.5200, 13.4050	(reprocessing recommended)	54 %	unclear

Cost table (billing basis)

Position	Unit	Quantity	Unit price	Sum
Initial processing	Document pass	200	0.15 EUR	30.00 EUR
Reprocessing (confidence optimization)	Document pass	7	0.15 EUR	1.05 EUR
Total net · 207 passes				31.05 EUR

Unit price 0.15 EUR per pass per document. Documents with low confidence pass through the AI pipeline multiple times (image AI localization, OCR, cross-check) until the desired confidence threshold is reached. Each pass is billed. Final invoicing is monthly per cost center.

Reprocessing recommendation (AI)

Local image AI reprocessing is recommended for 7 items:

IMG_4612.jpg, IMG_4709.jpg: EXIF coordinates 0.0 (camera defect). Visual localization via landmark detection is possible.
IMG_4613.jpg, IMG_4655.jpg: hits with confidence < 70 %. Image comparison against street-view embeddings can sharpen the result.
IMG_4801.jpg: 3 GPS jumps in EXIF, likely drift. Median filter recommended.
IMG_4823.jpg, IMG_4824.jpg: indoor shots with uncertain GPS. Manual placement via UI map view.

Download & actions

Report PDF

LATUS-CI · A4 · ~ 8 pages

ready

Content
Cover, KPIs, item excerpt, costs, AI recommendation

Size
~ 1.2 MB

Enrichment as ZIP

193 files · renamed via pattern

ready

Format
Pattern from step 02 applied

Size
~ 12 MB

Item table Excel

200 rows · all fields + confidence

ready

Sheets
Items · Stats · Costs

Size
~ 280 KB

Every deletion generates an audit proof (SHA-256 hash, timestamp, operator, run ID). The signed PDF proof is available for 90 days and documents the deletion GDPR-compliantly (Art. 17 + Art. 30). Proofs are then available in the "Audit proofs" tab.

Auto-cleanup: Delete data automatically after days

Run	Date	Input	Output	Total storage
Field photos Q2-2026 `run-20260511-1430`	2026-05-11	50 MB (200 items)	12 MB	62 MB
Telekom address points `run-20260511-0915`	2026-05-11	200 MB (1,500 items)	48 MB	248 MB
FTTH address check May `run-20260510-2200`	2026-05-10	8 MB (8 items)	2 MB	10 MB
Consulting audit Hesse `run-20260509-1730`	2026-05-09	120 MB (340 items)	22 MB	142 MB
Pilot test `run-20260508-1015`	2026-05-08	15 MB (25 items)	3 MB	18 MB
5 runs		393 MB input	87 MB output	480 MB total

Complete register of all deletion audit proofs generated so far. Every entry holds a SHA-256 hash, UTC timestamp, operator and is available as a signed PDF for 90 days (GDPR Art. 17 + Art. 30).

Run / dataset	Deleted on	Operator	SHA-256 (truncated)	GDPR
Pilot areas 2026-Q1 `run-20260503-1000`	2026-05-10 10:00 UTC	system (auto)	`sha256:7a3e1f...c92e`	Art. 17 (auto-cleanup)
Test import April `run-20260428-1530`	2026-05-09 08:14 UTC	stephan@latus-consulting.de	`sha256:b29d4f...0e21`	Art. 17 (manual)
Surveying test `run-20260415-0945`	2026-05-05 16:42 UTC	stephan@latus-consulting.de	`sha256:4c8e1a...88a1`	Art. 17 + Art. 30
Sample dataset Q1 `run-20260402-1200`	2026-05-02 12:00 UTC	system (auto)	`sha256:d7e290...4a01`	Art. 17 (auto-cleanup)

Signature: Ed25519 with the LATUS audit key (rotated quarterly). Retention 90 days, after that a hash index for 5 years on cold storage.

Geocoding AI

Upload

Processing mode & configuration

Run

Reports & admin

Upload

Local upload

Customer data room

Drop files here or click to select

New upload link

Active links 4

File sets 3

Audit history 12

Processing mode & configuration

Reverse geocoding

Forward geocoding

Address normalization

Run

Ready to start

Reports & admin

Current run

Run history 5

Delete

Audit proofs 4

Field photos Q2-2026

Summary (AI-generated)

KPIs

Distribution by federal state

Item details (excerpt)

Cost table (billing basis)

Reprocessing recommendation (AI)

Download & actions

Report PDF

Enrichment as ZIP

Item table Excel

Field photos Q2-2026

Telekom address points

FTTH address check May

Consulting audit Hesse

Pilot test