AI data annotation
Zipang runs video, image, text, and 3D-point-cloud annotation pods against your rubric. 5-gate screening funnel, weekly KPI dashboards, 7–14 day ramp, 30/90-day replacement guarantee. All-in USD 700–1,800 per annotator per month under PT Lima Cakar Bumi.
Zipang runs managed Indonesian AI-data-annotation pods for global AI startups, dataset providers, and autonomous-vehicle teams. Video, image, text, and 3D-point-cloud datasets are handled under the same 5-gate screening funnel and KPI dashboards that produced 432 deployed annotators, 208 in active production, 3.4M annotated tasks per month at 90%+ sustained accuracy for a 100+ hypermarket retail network in France, and 90%+ sustained accuracy on Transperfect Dataforce interpolation programs. Pricing is all-in USD 700–1,800 per annotator per month — base salary, PPh 21, BPJS Kesehatan, BPJS Ketenagakerjaan, THR, severance reserve, manager oversight, and the 30/90-day replacement guarantee are bundled under PT Lima Cakar Bumi as the legal employer of record. A pilot pod ramps from signed SOW to first productive shift in 7–14 days; full production teams of 50–200 annotators typically ramp in 3–6 weeks with phased onboarding, and the program is run against the buyer's rubric, not a generic label taxonomy.
432 deployed · 3.4M tasks / month · 90%+ accuracy · 7–14 day ramp · 30/90-day replacement guarantee.
Annotation modalities
The same 5-gate screening funnel backs video, image, text, and 3D-point-cloud annotation. The paid trial at the 5th gate is scored against the buyer's actual rubric — accuracy and throughput are validated before the production ramp.
Bounding boxes, polygon segmentation, action labels, interpolation between keyframes, and temporal event tagging. The same modality stack that produced 3.4M annotated video tasks per month at 90%+ accuracy for a 100+ hypermarket retail network in France.
Classification, object detection, semantic / instance segmentation, polygon and keypoint annotation. Rubric-scored, double-pass QA on tight-SLA programs.
NER, sentiment, intent, classification, span marking, and RLHF ranking. English and Bahasa Indonesia; second-language review pass for ambiguous cases.
Cuboid, segmentation, and tracking on autonomous-driving and robotics datasets. Trained annotators with autonomous-vehicle rubric experience.
How a pod ramps
The same 5-step operator flow that backs /employers/how-it-works, scoped for AI annotation work. The fifth step tracks accuracy, throughput, and review-cycle time against your rubric.
Buyer shares task taxonomy, label set, target accuracy, and target throughput. Zipang's annotation ops lead drafts a rubric and a paid-trial task that mirrors production load.
Candidates pass CV review, English proficiency, role-specific annotation quiz, structured video interview, and a paid trial scored against the buyer's own rubric.
Redacted profiles unlock in the client portal. Names stay hidden until the buyer approves and the candidate consents.
PT Lima Cakar Bumi issues the PKWT/PKWTT contract, registers BPJS Kesehatan and Ketenagakerjaan, and sets up the annotation platform access.
Accuracy, throughput, and review-cycle time are tracked weekly against the rubric. The 30/90-day replacement guarantee covers fit and non-performance.
Pricing
The regulatory floor is the 2026 Jakarta UMP (IDR 5,729,876, ~$342) plus 10.2% BPJS Ketenagakerjaan — that's ~$400/month all-in labor cost. The buyer invoice is the AI annotation role band on top.
IDR 5.729.876
2026 UMP (≈ $342 USD / month)
+ 10.2% BPJS Ketenagakerjaan (JHT 2% + JP 1% + JKK 0.24–1.74% + JKM 0.3%) = ~$400 / month all-in labor cost.
USD 700–1800/ annotator / month, all-in
Production proof
432 professionals
Deployed
208 in active production
In active production
3.4M production tasks / month
Production tasks / month
90%+ sustained accuracy in Transperfect Dataforce programs
Sustained accuracy (Transperfect Dataforce)
88%+ 12-month retention across the operator's book
12-month retention
7–14 days from signed SOW to first productive shift
Time to first shift
30-day replacement guarantee for fit and skill issues, 90-day for non-performance
Replacement guarantee
50–70% below US in-house equivalents for the same role
Vs US in-house
Named clients on the AI annotation funnel
AI annotation FAQ
Video annotation (bounding boxes, segmentation, action labels, interpolation), image annotation (classification, detection, segmentation, polygon), text annotation (NER, sentiment, intent, classification, RLHF ranking), and 3D-point-cloud / LiDAR annotation. The same 5-gate funnel backs every modality; the rubric is yours.
90%+ sustained accuracy in Transperfect Dataforce interpolation programs, and the same 5-gate funnel that produced 3.4M production tasks per month at 90%+ accuracy for a 100+ hypermarket retail network in France. Accuracy is tracked weekly against your rubric, not promised once at contract signing.
Most roles go from signed SOW to first productive shift in 7–14 days. The driver is the paid trial at the 5th gate of Zipang's screening funnel — you see accuracy numbers on your own workload before committing to a full ramp. Larger teams (50+ seats) typically ramp in 3–6 weeks with phased onboarding.
All-in USD 700–1,800 per annotator per month. The 2026 Jakarta UMP floor (IDR 5,729,876, ~$342) plus 10.2% BPJS Ketenagakerjaan lands the all-in labor cost at ~$400/month; the role band is the buyer's invoice on top. PPh 21, THR, severance, and year-end filings are bundled. No separate statutory add-ons.
Yes. PT Lima Cakar Bumi is the legal employer of record — every Zipang hire is on its payroll. The same entity handles PPh 21, BPJS Kesehatan, BPJS Ketenagakerjaan (JHT 2% + JP 1% + JKK 0.24–1.74% + JKM 0.3% = 10.2% all-in), THR, severance under UU 13/2003, and year-end SPT filing.
Yes. The pricing model is per agent and works the same way for a 5-annotator pilot and a 200-annotator program. Volume discounts apply at 50+ seats. The 30/90-day replacement guarantee covers fit and non-performance for every seat on the contract, not just the pilot.
A 30-day replacement guarantee covers fit and skill issues — if an annotator is not the right match in the first month, Zipang replaces them at no cost. A 90-day replacement guarantee covers non-performance — if accuracy or throughput KPIs are not met through the first quarter, Zipang replaces them too.
Free client registration gives you portal access to the redacted talent bench and a paid-trial rubric. Most pilot pods hit first productive shift in 7–14 days.
Reviewed by Yoseph Gratika, Founder & CEO (PT Lima Cakar Bumi). Last updated 2026-06-16. The 5-gate funnel, EOR layer, and production-proof numbers are the same operator assets used in the French retail AI annotation case study at /case-studies/french-retail-ai-annotation. Methodology and regulatory references are at /methodology.
