Moontask is a Web3 marketing intelligence platform built on Solana. Users complete tasks and earn token rewards via smart contracts on-chain. The challenge: build a system that could reliably separate genuine users from bots across a database of millions of wallets, and turn that engine into the backbone of a B2B verification API for the broader Web3 ecosystem.
Bot farms are a structural problem in Web3. When we analyzed Moontask's claim data, the pattern was unmistakable: organic activity followed by massive, coordinated non-human peaks. A bot farm had found the gaps: no time windows, no claim caps, no way to tell who was real.
Their wallet database held more than 2 million addresses. Only a small fraction had any social verification layer attached. The signal-to-noise ratio made it impossible to trust any reward distribution. They needed a way to score every wallet reliably, in real time and know with confidence who was human. And beyond that: they needed this engine to power a B2B product that other Web3 projects could plug into.
A combined scoring engine
We designed a two-signal system combining on-chain wallet behavior with off-chain social activity. Every wallet receives a score between 0 and 100. The model produced clear, validated separation between bot clusters and genuine users across two distinct cohorts drawn from Moontask's own production data, wallets from a high-registration period (bot-like behavior) versus a low-registration period (human-like behavior). The score distribution was distinct and actionable.
Wallet scoring via Helius API
The wallet score pulls on-chain signals directly from Solana: transaction count, wallet age, DeFi program participation, balance, and token diversity. A fresh, empty wallet, the classic bot signature, scores at the low end. We validated the separation across both cohorts; the distributions were distinct enough to drive reliable gating decisions without KYC.
Social scoring via Amazon Bedrock
The social score runs an Amazon Bedrock agent against Discord messages, analyzing for AI-generated content patterns: message diversity, activity cadence, and linguistic signals. We validated it against real human-written and AI-generated Discord messages. The agent not only caught obvious bot patterns — it flagged simple, repetitive messages as bot-like, demonstrating genuine sensitivity to non-human behavior rather than just surface-level keyword matching.
AWS-native serverless pipeline on Solana
Lambda functions host the Python scoring pipelines, write scores directly to RDS PostgreSQL, and connect via Discord webhooks for event triggering confirmed live during the demo. All infrastructure runs in AWS eu-central-1 and is provisioned with Terraform for clean handover and repeatability. The Solana-native architecture means verification happens at the transaction layer, not bolted on top of it
Working with Apex Lab was like having a co-founder who cared for the company as if it were their own, rather than treating us like just another business client. The infrastructure they built didn't just solve our bot problem, it became the engine behind our B2B product and directly accelerated our first enterprise customer pipeline. That kind of technical partnership is rare.
Grigory Mkrtchyan Founder & CEO
The PoC produced clear, validated separation on real production wallet data. The entire AWS infrastructure ran at approximately $60/month during the engagement. The production roadmap is defined: SQS-driven Lambda processing for scale, Helius API developer tier to unlock Sybil farm detection, WAF and rate limiting on public endpoints. At that specification, the system supports 2 million wallets per month at approximately $600/month.
The classification engine built during this engagement became the foundation for Moontask's B2B portal: a verification API that Web3 marketing agencies and token projects can integrate directly. The PoC didn't just validate the technology; it accelerated Moontask's first enterprise customer acquisition pipeline. Moontask is entering pre-seed fundraising with a working bot detection engine, a live B2B product path, and a proprietary dataset of more than 2 million classified Solana wallets as a defensible data moat.
Tell us more about yourself and what you're got in mind.
Prefer email? hello@apexlab.io
Or even a quick call?
We don’t care about fancy CVs or long interview processes. Provide some basic information, let’s have a quick chat, and then show us what you love doing the most: write some code and create a product.
