Procure
Saudi procurement intelligence
- Status
- Live indexing
- Dataset size
- 2.4M tender lines
- Last update
- 2026-03-07
AI Product Lab
oozr
Experiments in AI-native software, data moats, and automated products. oozr is built as a space-traversing, high-frequency signal in an age where noise scales faster than utility.
More information is not always better. The lab focus is filtering, ranking, and operationalizing what matters before entropy compounds.
Projects
Active experimentsRegulatedPlants
Global invasive species regulation dataset
- Status
- Data expansion
- Dataset size
- 13.1k regulations
- Last update
- 2026-03-06
CEFR.ai
Language difficulty modelling
- Status
- Model tuning
- Dataset size
- 96k labeled passages
- Last update
- 2026-03-08
Live Metrics
Auto refreshTenders indexed
184,620
Updated 2026-03-08 09:22
Plant regulations collected
13,112
Updated 2026-03-08 09:18
CEFR texts analyzed
96,340
Updated 2026-03-08 09:24
Experiments completed
57
Updated 2026-03-08 09:10
Usage Signals
Events / daycefr.ai texts analyzed
1,260
0 users today
824.4% 7d
procure searches
730
0 users today
770.1% 7d
regulated plants lookups
418
0 users today
767.3% 7d
| Growth metric | Latest | 24h delta | 7d delta | 7d growth |
|---|---|---|---|---|
| tenders indexed | 184,620 | 14,380 | 14,380 | 8.4% |
| plant regulations collected | 13,112 | 1,072 | 1,072 | 8.9% |
| cefr texts analyzed | 96,340 | 7,940 | 7,940 | 9.0% |
Combined mrr 7d change: $830.00 (12.8%)
Sync health
cefr-ai
no sync runs yet
procure
no sync runs yet
regulatedplants
no sync runs yet
Revenue + Spend
Open finance ->Monthly revenue
$7,330.00
Monthly spend
$163.84
Net monthly
$7,166.16
Shared tools burn
$78.25
| Project | Revenue | Spend | Net |
|---|---|---|---|
| Procure | $4,200.00 | $68.00 | $4,132.00 |
| RegulatedPlants | $1,850.00 | $51.75 | $1,798.25 |
| CEFR.ai | $1,280.00 | $44.08 | $1,235.92 |
Build Log
ChronologicalDate
2026-03-08
Experiment
Prompt-augmented CEFR classification on noisy learner text
Result
Reduced borderline A2/B1 misclassification by 12%.
Code
Date
2026-03-07
Experiment
Tender deduplication with supplier + title fingerprinting
Result
Removed 8.4% duplicate notices from weekly ingest.
Code
Date
2026-03-06
Experiment
Country-level rule extraction using constrained parsing
Result
Lifted extraction precision to 0.93 on validation set.
Code
Date
2026-03-05
Experiment
Supplier intent scoring from historical award signals
Result
Top-decile lead precision improved from 0.41 to 0.58.
Code
N/A
Date
2026-03-03
Experiment
Lexical rarity feature ablation for CEFR bands
Result
Feature had high impact on B2/C1 separation; kept in stack.
Code
N/A
Date
2026-02-28
Experiment
Crawl retry policy for unstable ministry portals
Result
Collection reliability rose from 78% to 95% weekly.
Code