Self-taught engineer building deployed on-chain systems and production ML pipelines. I ship full-stack: Solidity contracts on Base mainnet, ML strategies running against live exchanges, off-chain orchestrators, and the infrastructure that holds it together.
I'm not a generalist trying to look specialized — I'm a specialist who happens to span the full stack of what crypto + ML systems require.
Three ERC-20 tokens deployed and verified on Base (chainId 8453):
- WealthAI (WAI) — 420.69 billion supply with Aerodrome auto-liquidity, 7-bucket fee distribution (burn / liquidity / marketing / reflection / buyback / staking / treasury), passive reflection mechanics, timelocked admin operations, and post-launch auto-throttle.
0xb7D17C61... - VIBE — 1 billion supply ERC-20, Solidity 0.8.28, source verified.
0x8A09a968... - GIGA — ERC-20 deployed on Base.
0x5E2a88E5...
Multi-platform listing automation — single-command toolkit that builds Trust Wallet asset submissions, 1inch tokenlist entries, and submits to GeckoTerminal, DexScreener, CoinGecko, CoinMarketCap, and DexTools. Reduces a multi-day manual launch to a single script invocation.
Eight upstream pull requests to trustwallet/assets across three tokens, all merge-ready with proper logo specs and metadata.
- LightGBM classification strategy running on Kraken via Freqtrade — 17 pairs on 4h timeframe, 77 features, single gradient-boosted classifier (not an ensemble). Trained on multi-year 4h OHLCV data; retraining pipeline emits feather-format datasets to Freqtrade's data directory. Currently in dry-run while accumulating live forward-test data — the committed equity curve shows the real result, including drawdown, rather than a cherry-picked number.
- Custom ATR-scaled stop-loss layered on top of the standard Freqtrade stoploss interface, with regime-aware threshold adjustments (risk-on / risk-off / neutral).
- Honest backtest discipline — 139-trade closed-position diagnostic with regime breakdown and lookahead-bias checks. When a v17 backtest showed suspicious results, I flagged it and rolled back to v16 rather than ship it. I removed an in-sample AUC figure from an earlier version of this page once I couldn't validate it out-of-sample — if a number isn't verified, it doesn't go on the résumé.
- Forward-walk paper trading simulator (Solana / Jupiter) — real entry/exit tracking, ATR stops, SQLite persistence, labeled decision logging for future modeling.
- Real-time market scanner monitoring 10 USDT pairs every 5 minutes with BTC dominance tracking (CoinGecko API), 30-day rolling cross-pair correlations, AI-generated alert summaries, and a web dashboard backed by FastAPI.
- vibe-cockpit — a local telemetry dashboard that unifies multiple trading bots: live PnL, per-bot drill-down (equity, regime, latency, signals), a filterable trade feed, and a raw SQLite explorer. Schema-agnostic ingest with automatic lamports/SOL unit detection.
Composed a flash-loan arbitrage architecture for Arbitrum: Aave V3 flash loans, dual-DEX routing (Uniswap V3 + Sushiswap), yield aggregation (Yearn + Beefy), Chainlink oracle integration, and TimelockController governance — in a single contract. The off-chain orchestrator wires up Web3.py, AWS KMS for envelope-encrypted key management, and supporting tooling for scaling and monitoring. I also explored Qiskit for path-optimization research, but that piece is exploratory, not a working optimizer. Research prototype, documented honestly as pre-audit and pre-deployment — it has not been audited or run against real funds.
I build things that run. I don't ship code I haven't tested against real data — and when something doesn't work, I don't dress it up to look like it does. That discipline is why my backtests catch lookahead bias, my README files document what's actually deployed versus what's scaffolded, and my contracts get verified on Basescan instead of staying anonymous.
I learn by reading source code, building from scratch, breaking things, and fixing them. I'm comfortable in Solidity, Python, JavaScript, and Bash. I work end-to-end: from contract design to deployment scripts to off-chain orchestration to dashboards.
- Smart Contracts: Solidity 0.8.x, Hardhat, OpenZeppelin v5, Aave V3, Uniswap V3, Aerodrome, Chainlink, ethers.js v6
- ML / Data: Python, LightGBM, scikit-learn, pandas, numpy, custom backtest pipelines
- Trading: Freqtrade, Kraken & Binance APIs, custom strategies, technical indicators, regime detection
- Off-chain: FastAPI, Web3.py, AWS boto3, Telegram bots
- Infrastructure: Docker, Redis, PostgreSQL, GitHub Actions, Linux (macOS daily driver)
Based in St. John's, Newfoundland (Atlantic time, UTC-3). Open to remote engineering roles in crypto, ML, or full-stack — particularly roles where shipping production systems matters more than credentials.
Twitter: @OfficialGIGA