Frequently Asked Questions

Everything you need to know about MicroDC.ai

Information for CEOs, CFOs, and business decision-makers evaluating MicroDC.ai for their organization.

1 Company & Platform Overview

AI inference is the process of using a trained AI model to generate outputs from new inputs. Think of it as "running" the AI to get answers, summaries, or predictions.

Simple Analogy:

If training an AI is like teaching a student (expensive, time-consuming), then inference is like asking that student questions after they've graduated (quick, but still requires resources).

Examples of AI inference:

  • Asking ChatGPT a question and receiving an answer
  • Having an AI summarize a 50-page document
  • Using AI to classify thousands of customer support tickets
  • Generating product descriptions from catalog data
  • Extracting key information from contracts or invoices

Why it's expensive: AI inference requires powerful GPU hardware. Every time you use ChatGPT, Claude, or similar services, GPUs are working behind the scenes. At scale, these costs add up quickly - which is exactly the problem MicroDC solves.

Bottom line: If your business needs to process text, documents, or data using AI at scale, you're paying for inference. MicroDC helps you pay up to 90% less.

MicroDC.ai is a distributed AI inference platform that operates as a two-sided marketplace, connecting businesses that need AI compute power with individuals and organizations that have idle GPU resources.

The Problem: AI inference costs are skyrocketing. Major cloud providers charge premium prices for GPU instances, and building internal infrastructure requires significant capital expenditure and technical expertise.

Our Solution: We aggregate distributed GPU resources from around the world, providing enterprise-grade AI inference at up to 90% lower cost than traditional cloud providers, with no infrastructure management required.

MicroDC was founded by Ray Sisson (CEO) and Jeffrey Rivero (CTO), with Neil Quarterman serving as CFO.

  • Ray Sisson (CEO): MBA from University of Chicago Booth, JD from Georgetown. Extensive experience in transportation, infrastructure, and corporate board leadership.
  • Jeffrey Rivero (CTO): 25+ years in cybersecurity and AI/ML. CISSP certified, built platforms handling 1M+ events/second, developed AI solutions for the US military.
  • Neil Quarterman (CFO): Yale and Georgia Tech MBA. Experience managing financials through 140% growth and M&A across 11 entities in 8 countries.
Factor MicroDC Major Cloud Providers
Business Model Marketplace (asset-light) Infrastructure owner (asset-heavy)
Cost Savings Up to 90% lower Premium pricing
Processing Model Asynchronous batch Real-time (expensive)
Minimum Commitment Pay-per-use, no minimum Often requires commitments
Setup Time Minutes Hours to days

2 Business Value & ROI

Organizations typically see 50-90% cost reduction compared to traditional cloud GPU instances:

  • Batch Processing: Our asynchronous model eliminates idle compute charges
  • No Infrastructure Overhead: Zero cost for DevOps, maintenance, or scaling
  • Pay-Per-Use: Only pay for actual tokens processed, not reserved capacity
  • No Minimum Commitment: Scale from $10 to $10,000/month seamlessly

Example: A document processing company processing 100,000 documents/month saved 78% by switching from dedicated GPU instances to MicroDC's batch processing.

MicroDC excels at workloads that don't require sub-second real-time responses:

Ideal Use Cases

  • Document summarization & analysis
  • Content generation at scale
  • Data enrichment & classification
  • Research & batch analytics
  • Overnight processing jobs
  • Email/report generation

Consider Alternatives For

  • Real-time chatbots (<1s response)
  • Interactive user-facing AI
  • Streaming responses
  • Mission-critical real-time systems

We use a simple, transparent credit-based system:

  • Credits: 1 credit = $0.01 USD
  • Welcome Bonus: 10 free credits for new accounts
  • No Hidden Fees: No setup, monthly, or platform fees
  • Per-Job Pricing: Based on model complexity and tokens processed

Pricing Tiers:

  • Economy: Lower cost, flexible timing
  • Standard: Balanced cost and performance
  • Premium: Priority processing

3 Security & Compliance

Security is foundational to our architecture. Our CTO is CISSP-certified with 25+ years in cybersecurity:

  • End-to-End Encryption: Data encrypted in transit (TLS 1.3) and at rest
  • Worker Isolation: Workers cannot see customer identities or access other jobs
  • Data Minimization: Only job payloads are sent to workers; no metadata leakage
  • Automatic Deletion: Job data purged within configurable retention periods
  • Audit Trails: Complete logging of all transactions and access
  • JWT Authentication: Secure token-based authentication with short-lived access tokens

Current Security Posture:

  • Security-first architecture designed by CISSP-certified leadership
  • Regular security audits and penetration testing
  • Secure development lifecycle practices

Planned Certifications (Roadmap):

  • SOC 2 Type II
  • ISO 27001
  • GDPR compliance documentation

Contact us for detailed security documentation and our current compliance status.

We recommend the following approach for sensitive data:

Best Practices:

  • Anonymize or pseudonymize data before processing
  • Remove PII from job payloads when possible
  • Use our data retention settings to minimize exposure
  • Consider our upcoming private worker network for enhanced control

For HIPAA, PCI-DSS, or other regulated workloads, please contact our sales team to discuss dedicated infrastructure options.

4 Partnership & Enterprise Support

Yes, we offer customized enterprise packages including:

  • Volume Discounts: Significant savings for committed usage
  • Priority Processing: Dedicated capacity and faster job completion
  • SLA Guarantees: Uptime and performance commitments
  • Dedicated Support: Named account manager and technical support
  • Custom Integration: API customizations and on-premise options
  • Invoice Billing: NET-30 payment terms for qualified customers

Contact sales@microdc.ai to discuss enterprise options.

Platform Availability: 99.9% API uptime target

Job Processing SLA: Varies by priority tier

  • Economy: Best-effort, typically <4 hours
  • Standard: 95% completed within 1 hour
  • Premium: 99% completed within 15 minutes

Enterprise customers receive customized SLAs with credits for missed targets.

Organizations with idle GPU capacity can monetize their infrastructure:

  • Revenue Share: Workers receive 70% of job revenue (industry-leading)
  • Flexible Participation: Connect resources on your schedule
  • Automatic Job Distribution: Our scheduler optimizes job assignment
  • Transparent Earnings: Real-time dashboard for earnings and payouts
  • Multiple Payout Options: Credits, PayPal, or bank transfer

Contact us to discuss becoming a GPU provider partner.

5 Future Roadmap & Blockchain Integration

We are developing blockchain integration to enhance transparency, trust, and decentralized payments:

Planned Blockchain Features:

  • Transparent Job Verification: Immutable record of completed work and payments
  • Smart Contract Payments: Automated, trustless payments to GPU providers
  • Decentralized Reputation: On-chain worker reputation scores
  • Token-Based Incentives: Potential utility token for platform participation
  • Cross-Border Payments: Simplified international worker payouts via crypto

Blockchain integration is on our roadmap. Traditional payment methods will remain available for organizations preferring conventional finance.

  • Auditable Compute: Cryptographic proof of work completed for compliance
  • Cost Transparency: Verifiable pricing with no hidden markups
  • Dispute Resolution: Smart contract arbitration for contested jobs
  • Global Settlement: Instant international payments without banking delays
  • Data Integrity: Tamper-proof logs of all platform interactions
Now

Core Platform

Batch inference, 100+ models, credit system, web dashboard

Next

Enhanced Capabilities

Streaming responses, custom model hosting, enhanced analytics

Future

Blockchain & Scale

On-chain verification, token payments, global expansion

Technical details for engineers and developers building with MicroDC.ai.

1 Technical Architecture

MicroDC uses a three-component distributed architecture:

Client (SDK/API)     Server (Hub)         Workers (Compute)
      |                   |                     |
      |  Submit Job       |                     |
      |------------------>|                     |
      |                   |  Queue Job          |
      |                   |-------------------->|
      |                   |                     |
      |                   |  Poll for Jobs      |
      |                   |<--------------------|
      |                   |                     |
      |                   |  Return Results     |
      |                   |<--------------------|
      |  Get Results      |                     |
      |<------------------|                     |
                                        

Key Components:

  • Server: API backend, PostgreSQL database, job scheduler
  • Workers: Distributed GPU nodes running inference engines
  • Client: Python SDK and REST API for job submission

Backend

  • API
  • PostgreSQL (primary database)
  • JWT authentication
  • containerization

Frontend

  • HTMX for dynamic UI
  • Alpine.js for reactivity
  • Tailwind CSS styling
  • Jinja2 templating

Workers

  • Python worker runtime
  • vLLM / llama.cpp inference
  • CUDA/ROCm GPU support
  • Automatic model management

Infrastructure

  • Docker Compose (dev)
  • Kubernetes (production)
  • Nginx reverse proxy
  • Redis caching (planned)

The scheduler uses intelligent matching to optimize job distribution:

  1. Job Submission: Client submits job with model requirements
  2. Capability Matching: Scheduler finds workers with required GPU/VRAM
  3. Priority Queuing: Jobs ordered by priority tier and submission time
  4. Worker Selection: Best-fit algorithm considers worker performance history
  5. Failure Handling: Failed jobs rescheduled to different workers

Smart Retry: If a job fails on a worker, it will not be reassigned to the same worker to prevent repeated failures.

2 API & SDK Integration

Install the SDK and submit jobs with just a few lines:

# Install
pip install git+https://gitlab.com/microdc/python-client.git

# Submit a job
from microdc import Client, LLMCall

client = Client(api_key="mDC_your_api_key")

# Create an LLM job
job = LLMCall(model="llama3.3")
job.add_user_message("Explain quantum computing in simple terms")

# Submit and wait for result
job_id = client.send_job(job)
result = client.wait_for_job(job_id)
print(result.output)
                                        

View full SDK documentation on GitLab

Core API endpoints (base URL: https://console.microdc.ai/api/v1):

Method Endpoint Description
POST /jobs Submit a new job
GET /jobs/{job_id} Get job status/result
GET /jobs List your jobs
GET /models List available models
GET /account/balance Check credit balance
DELETE /jobs/{job_id} Cancel a pending job

Authentication: Include header Authorization: Bearer mDC_your_api_key

LLMCall

Text generation and chat completions

  • Chat/conversation format
  • System prompts supported
  • Temperature/top_p controls

EmbedCall

Text embeddings for search/RAG

  • Batch embedding support
  • Multiple embedding models
  • Vector output formats

ImageGeneration

AI image creation

  • Stable Diffusion models
  • Custom dimensions
  • Negative prompts

CustomJob

Flexible custom workloads

  • Custom model support
  • Arbitrary payloads
  • Advanced configurations

You can receive job completion notifications via webhooks:

# Submit job with callback URL
job = LLMCall(model="llama3.3")
job.add_user_message("Process this document...")

job_id = client.send_job(
    job,
    callback_url="https://your-api.com/webhook/microdc"
)

# Your webhook will receive:
# POST https://your-api.com/webhook/microdc
# {
#   "job_id": "uuid",
#   "status": "completed",
#   "result": { ... }
# }
                                        

Alternatively, use polling with client.wait_for_job(job_id)

3 Models & Performance

We support 100+ open-source models including:

LLM Models

  • Llama 3.3 (8B, 70B)
  • Mistral (7B, Mixtral)
  • Qwen 2.5
  • DeepSeek
  • Phi-3
  • Gemma 2

Embedding Models

  • BGE (various sizes)
  • E5
  • GTE
  • Nomic Embed

Image Models

  • Stable Diffusion XL
  • SDXL Turbo
  • Flux

View complete model list with pricing

Yes! Custom model hosting is available:

  • Model Upload: Upload your fine-tuned model weights
  • Validation: Automatic testing and validation
  • Deployment: Deployed to compatible workers
  • Version Control: Manage multiple model versions
  • Private Access: Models only accessible to your account

Contact us for custom model hosting requirements and pricing.

Job completion times depend on several factors:

Factor Impact
Priority Tier Premium < Standard < Economy
Model Size 7B faster than 70B
Token Count More tokens = longer processing
Queue Depth Varies by network load

Typical Times: Most jobs complete within 30 seconds to 5 minutes. Premium tier averages under 1 minute.

4 Security & Authentication

We use API keys for SDK/API authentication:

# Environment variable (recommended)
export MICRODC_API_KEY="mDC_your_api_key_here"

# In code
client = Client()  # Auto-reads from env

# Or explicitly
client = Client(api_key="mDC_...")

# HTTP Header
Authorization: Bearer mDC_your_api_key
                                        

Best Practice: Never commit API keys to version control. Use environment variables or secrets management.

  • In Transit: TLS 1.3 encryption for all API calls
  • At Rest: AES-256 encryption for stored data
  • Worker Isolation: Jobs processed in isolated environments
  • Data Minimization: Only necessary data sent to workers
  • Automatic Purging: Results deleted after configurable retention
  • No Logging of Payloads: Job content not logged server-side

Default rate limits (can be increased for enterprise):

Job submissions 100/minute
Status checks 300/minute
Concurrent jobs 50 (default)
Burst allowance 2x for 10 seconds

Rate limit headers included in responses: X-RateLimit-Remaining

5 Becoming a GPU Provider (Worker)

Minimum requirements:

Component Minimum Recommended
GPU NVIDIA GTX 1060 6GB RTX 3080+ / A100
VRAM 6GB 16GB+
RAM 16GB 32GB+
Storage 100GB SSD 500GB+ NVMe
Internet 50 Mbps 100+ Mbps

AMD GPUs (ROCm) also supported. CPU-only workers can handle smaller models.

Quick setup (5 minutes):

# 1. Clone the worker repository
git clone https://gitlab.com/microdc/worker.git
cd worker

# 2. Install dependencies
pip install -r requirements.txt

# 3. Configure with your registration token
export MICRODC_WORKER_TOKEN="mdc_wrk_your_token"

# 4. Start the worker
python worker.py --models llama3-8b,mistral-7b

# Worker will automatically:
# - Register with the server
# - Download required models
# - Start polling for jobs
                                        

View detailed worker setup guide

Workers receive 70% of the job price (industry-leading):

Example Calculation:

  • Job Price: $0.10
  • Worker Share (70%): $0.07
  • Platform Fee (30%): $0.03

Factors affecting earnings:

  • Model complexity (larger models pay more)
  • Job volume and availability
  • Worker performance/reliability score
  • Geographic location (latency preferences)

6 Blockchain Integration (Coming Soon)

We're evaluating several blockchain platforms for our integration:

Under Evaluation

  • Ethereum L2s (Arbitrum, Optimism)
  • Solana (high throughput)
  • Polygon (low fees)
  • Custom app-chain (Cosmos SDK)

Selection Criteria

  • Low transaction fees
  • High throughput
  • Developer ecosystem
  • Regulatory clarity

Final blockchain selection will be announced as we approach launch.

Smart contracts will automate several platform functions:

Payment Escrow

Job payment locked in escrow, released to worker on verified completion

Proof of Work

Cryptographic verification that inference was performed correctly

Reputation System

On-chain reputation scores for workers based on performance

Dispute Resolution

Decentralized arbitration for contested job completions

We're exploring a utility token model with the following potential features:

  • Payment Currency: Alternative to fiat credits for job payments
  • Staking: Workers stake tokens as collateral for quality assurance
  • Governance: Token holders vote on platform parameters
  • Fee Discounts: Token payments may receive reduced platform fees
  • Rewards: Early adopters and high-performing workers earn bonus tokens

Note: Token economics are under development. Traditional fiat payments will always be supported alongside any token implementation.

Planned hybrid architecture:

Off-Chain (Speed)           On-Chain (Trust)
       |                          |
 Job Submission                   |
       |                          |
 Job Execution                    |
       |                          |
 Result Delivery                  |
       |                          |
       +--- Settlement ---------->| Payment Record
       |                          | Proof of Work
       +--- Disputes ------------>| Arbitration
       |                          | Reputation Update
                                        

Key Design Principles:

  • Off-chain for speed-critical operations (job processing)
  • On-chain for trust-critical operations (payments, disputes)
  • Batch settlement to minimize gas costs
  • Optimistic verification with challenge period

Blockchain integration is on our roadmap with a phased approach:

Phase 1

Crypto Payments

Accept stablecoin payments (USDC, USDT)

Phase 2

On-Chain Verification

Job completion proofs on-chain

Phase 3

Full Decentralization

Token launch, staking, governance

Sign up for our newsletter to stay updated on blockchain developments.

Still Have Questions?

Our team is here to help. Reach out for personalized assistance with your specific use case.