Live · Base Mainnet · USDC Intelligence

Senior Data Architect who
transforms raw, complex data into actionable intelligence — self-taught

I design and build end-to-end systems that turn raw blockchain activity into actionable financial intelligence. OmniSight ingests Base Mainnet USDC flows, decodes EVM payloads, and scores wallet behaviour with ML — production-grade, latency-aware, and API-first.

Focus: Real-time Web3 risk & compliance systems

Stack: Airflow · web3.py · PostgreSQL · FastAPI · Scikit-Learn · OCI

Overview

Why OmniSight matters

Web3 moves fast and ledger data is noisy. OmniSight turns raw USDC transfers into structured, ML-ready intelligence for compliance, fraud detection, and institutional reporting.

Core capabilities

Autonomous ingestion (Airflow) with 120s cadence
EVM hex decoding and USDC normalisation via web3.py
Partitioned PostgreSQL fact tables for scale
Isolation Forest ML anomaly scoring (RobustScaler)
FastAPI gateway — authenticated + public endpoints

Real-time

ML-driven

API-first

Production

Flagship project

OmniSight Web3 Data Engine

A production-grade intelligence engine built entirely from scratch — no bootcamp, no team, no starter template. Ingests Base Mainnet USDC flows, decodes them into partitioned fact tables, and scores wallet behaviour with a trained ML model exposed via a documented REST API.

System snapshot

120s ingestion cadence via Airflow pulling JSON-RPC event logs
EVM decoder → typed analytics fields (wallet, amount, block)
Partitioned tables — usdc_transfers_era_47m, era_48m, growing
RobustScaler + Isolation Forest — trained on 6,079 real wallet profiles
FastAPI + Nginx + TLS — live at ericdiamason.tech/docs
Oracle Cloud Infrastructure — always-on, zero cost (Always Free tier)

Business outcomes

Compliance-grade monitoring for exchanges and institutions
Real-time alerts for high-value movements ≥$50,000 USD
ML wallet risk scoring in under 200ms per request
Full audit trail: block number, tx hash, ingestion timestamp

Market context

The blockchain analytics market is projected to reach $18.7 billion by 2034 (22% CAGR). OmniSight demonstrates production-level capability in the space occupied by Chainalysis, Elliptic, and TRM Labs — built independently, from scratch, on a free cloud tier.

Compliance

Fraud detection

AML

Institutional

Public API

Live, documented endpoints. Try the wallet risk scorer below in the Live Metrics section.

GET /api/v1/metrics/whale-alerts GET /api/v1/public/wallet-risk?wallet_address=0x… View full API docs →

GitHub

Open source. Production code, clean secrets management, documented architecture.

github.com/ericdiamason/omnisight →

Architecture

From ledger to risk score in one coherent pipeline

Eight stages, zero gaps. Every layer is designed, deployed, and maintained by one engineer.

Pipeline

1 · Base Mainnet (JSON-RPC via Alchemy)

↓ eth_getLogs · USDC Transfer events

2 · Airflow DAG (120s schedule · max_active_runs=1)

↓ Incremental checkpoint · ON CONFLICT idempotency

3 · web3.py EVM decoder (32-byte topics → wallet addresses)

↓ 6-decimal USDC normalisation

4 · PostgreSQL partitioned tables (block-range partitions)

↓ RobustScaler feature normalisation

5 · Isolation Forest ML model (6,079 wallet profiles)

↓ asyncpg connection pool · Pydantic schemas

6 · FastAPI REST gateway (authenticated + public endpoints)

↓ TLS · HTTPS · CORS locked to domain

7 · Nginx reverse proxy · Let's Encrypt TLS

↓ systemd managed · survives reboots

8 · Oracle Cloud Infrastructure (Always Free · always-on)

Key design decisions

Partitioned tables: block-range partitioning for query performance at scale
RobustScaler: handles extreme whale outliers without discarding signal
ON CONFLICT idempotency: pipeline is safe to re-run against any block range
asyncpg pool: persistent connections, async I/O, handles concurrent requests
Secrets via EnvironmentFile: zero credentials in source code or git history
Rate-limited public endpoint: open access with abuse protection

Operational facts

214,000+ clean records, zero duplicates enforced by DB constraint
Model version v20260617 — trained on real Base Mainnet data
API survives server reboots via systemd service
Log rotation configured — no unbounded disk growth

Live Metrics

Real data. Real pipeline. Right now.

Every number below is pulled live from the OmniSight API — no mock data, no static fixtures.

Whale transfers (sample)

—

≥$50K · last 25

Total volume (sample)

—

USD

Largest transfer

—

USD

Ingestion latency

—

seconds chain→DB

Whale alerts — live feed

—

Block	Tx	Sender	Receiver	Amount (USD)	Ingested
Loading whale alerts…

Wallet Risk Explorer

Enter any wallet address to run it through the live Isolation Forest ML model. Powered by real Base Mainnet data.

Enter a wallet address above to score it.

Operational health

Ingestion: —

Model: —

Who I work with

Teams that benefit from OmniSight

I partner with Web3 infrastructure teams, fintech and compliance groups, exchanges, and analytics platforms to deliver production-grade telemetry and ML-driven risk intelligence.

Exchanges & Custodians

Real-time monitoring for large movements, counterparty tracing, and compliance workflows.

Compliance & AML

Automated alerts, ML risk tiers, and evidence-ready transaction trails for investigations.

Fintech & Analytics

Operational dashboards, institutional reporting, and data products built on reliable pipelines.

Contact

Let's build something that matters

If you need architecture, data engineering, or ML for Web3 risk and compliance, reach out. I consult, partner, and build production systems end-to-end.

Engagements I take

Proof-of-concept to production pipelines
ML risk scoring and model operationalisation
API design and integration for analytics teams
Architecture reviews and performance tuning
Web3 data infrastructure from scratch

Prefer a direct line?

admin@ericdiamason.tech LinkedIn → GitHub →