Skip to main content

Work

Real systems.
Sanitised for NDA.

Selected algorithmic-trading builds, sanitised for NDA.

Order bookBTC/USDT
42,180spread 10
  • bidsize
  • 42,1750.51
  • 42,1701.16
  • 42,1650.45
  • 42,1600.51
  • 42,1550.98
  • 42,1500.97
  • 42,1450.48
  • asksize
  • 42,1851.34
  • 42,1900.68
  • 42,1950.67
  • 42,2000.77
  • 42,2050.56
  • 42,2101.09
  • 42,2150.71
Inventory skew · −0.12Hedge · live

01 / Crypto · LATAM prop desk

Cross-exchange market-making bot

Reduced manual intervention per shift from ~4 hours to ~25 minutes.

NautilusTrader-based market maker across three exchanges with inventory-skew logic and real-time hedging. Reduced manual intervention per shift from ~4h to ~25min. Live across 18 trading pairs. NDA — 2025.

  • NautilusTrader
  • Python
  • Rust
  • Redis
  • Grafana
Discuss a similar build
FIX 4.4 · message flowNewOrderSingle
Client
Trader UI
FIX Gateway
4.4 session
OMS
Risk + state
Venue
Exchange
→ 8=FIX.4.4|35=D|49=CLIENT|56=BROKER|55=AAPL|54=1|38=100|40=2|44=190.45
← 8=FIX.4.4|35=8|39=0|150=0|17=00001|11=ord-7421
← 8=FIX.4.4|35=8|39=2|150=F|17=00002|31=190.43|32=100

02 / Equities · Argentine broker

FIX 4.4 connectivity layer

Reconnection logic, order-state machine, trade-reporting hooks.

FIX 4.4 gateway and OMS integration for a regulated broker, with reconnection logic, order-state machine and trade-reporting hooks. NDA — 2024.

  • FIX 4.4
  • Python
  • Postgres
  • Sentry
  • Datadog
Discuss a similar build
XGBoost · feature importanceregime classifier
  • VOLATILITY_5D0.92
  • REGIME_HMM0.78
  • BTC_CORR0.64
  • VOL_PROFILE0.55
  • RSI_140.41
  • MACRO_RATE0.28

Walk-forward AUC 0.71 · feature-drift dashboard live

03 / Quant research · US prop firm

XGBoost regime-classification signal

Walk-forward validated XGBoost feeding position-sizing on a discretionary futures book.

Walk-forward validated XGBoost classifier feeding position-sizing on a discretionary futures book. Drawdown bands and feature-drift dashboard delivered alongside. NDA — 2025.

  • XGBoost
  • pandas
  • scikit-learn
  • Optuna
  • Grafana
Discuss a similar build

More case studies ship every quarter.

More case studies coming as engagements close — Federico publishes new ones quarterly. Subscribe to /writing to see them first.