> For the complete documentation index, see [llms.txt](https://perle.gitbook.io/perle-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://perle.gitbook.io/perle-docs/platform-overview/system-architecture.md).
# System Architecture
How Perle Labs coordinates contributors, verification, rewards, and on-chain provenance.
The architecture consists of four coordinated layers:
### Contributor Layer: Human Expertise as Input
This is where contributors interact with Perle:
* **Wallet / Google / Email Authentication**
Contributors join using a decentralized or Web2 login method.
* **Training & Qualification**
Educational tasks verify the contributor’s skill before unlocking higher-value tasks.
* **Task Execution**
Contributors provide annotations, evaluations, and validations across text, image, audio, and video.
* **Reputation Building**
Accuracy and consistency update their reputation score, tier, and badges.
**Key Output:**
Validated human intelligence enters the Perle system as *structured, reviewable annotations.*
### Data Workflow Layer: Verification & Scoring Engine
Once a contributor submits work, the Perle verification pipeline begins:
* **Consensus Validation**
Multiple contributors perform the same task; the system measures alignment.
* **Gold-Standard Comparison**
Tasks with known correct answers validate accuracy.
* **Reviewer Checks (Coming Soon)**
High-tier contributors review and validate others’ work.
* **Reputation & Tier Updates**
Contributor accuracy dynamically adjusts their tier (Novice → Researcher → …).
* **Reward Scoring**
Points are assigned based on:
* correctness
* complexity
* streak multiplier
* referral multipliers
**Key Output:**
High-quality, validated annotation data + reputation metadata.
### On-Chain Provenance Layer (Solana)
Perle uses Solana for transparency and auditability, not for storing raw data.
**What goes on-chain:**
* **Contribution Records**
Proof that a contributor completed a task.
* **Reputation Snapshots**
Tier, accuracy, and performance proofs.
* **Reward Events (Future)**
When token conversion is enabled, reward claims will be cryptographically recorded.
**What does *****NOT***** go on-chain:**
* Raw data
* User personal information
* Annotation content itself
**Key Output:**
Tamper-proof history of who contributed what and when — essential for enterprise audit requirements.
### Reward & Economy Layer
The reward layer governs how contributors earn and how rewards will later be tokenized.
**Current State (Beta):**
* Contributors earn points for all tasks and activities.
* Points determine badges, streak multipliers, leaderboard position, and tier advancement.
**Future State (Token Launch):**
* A contribution-based token claim system will convert points into on-chain rewards.
* On-chain reputation will integrate directly with contributor wallets.
* Enterprise demand for expert labeling will influence reward supply and distribution.
**Key Output:**
A transparent, contribution-driven incentive system.
