Layer L1
Core engine
Strategy and execution logic on the hot path.
- C++17
- C++20
- Low-latency systems
- Multithreading and lock-free concurrency
- Tick-to-trade optimization
01 // THE STACK
The stack
Modern C++ on the hot path, proven protocols on the wire, and cloud compute for research. Every layer is picked for speed you can measure and behavior you can monitor.
Layer L2
Connectivity
Market data in, orders out, client accounts mirrored.
- FIX protocol
- TCP/UDP and multicast market data
- MetaTrader 5 (MT5)
Layer L3
Research and validation
Evidence first. Nothing goes live on a single backtest.
- Parallel backtesting
- Walk-forward and Monte Carlo validation
Layer L4
Observability and cloud
Live telemetry and elastic research compute.
- Prometheus
- Grafana
- AWS
02 // RESEARCH TO PRODUCTION
One team, one codebase, research to production
The engineers who write the research also ship the C++ and watch it trade. No hand-offs, and no rewrite between a research notebook and the engine.
01 · Build
Research into C++
Ideas become hypotheses. Hypotheses become production code.
- Research ideas framed as hypotheses with explicit entry, exit and risk rules
- Strategy logic implemented in C++17/20, one code path from research to production
- Risk limits written into the engine, not bolted on afterwards
- Version-controlled, reviewed code for every strategy change
Outputcompiled strategy core
02 · Test
Evidence before go-live
Every strategy has to survive data it has never seen.
- Parallel backtesting across instruments, parameters and market regimes
- Walk-forward analysis and out-of-sample holdouts to catch overfitting
- Monte Carlo resampling to stress drawdowns and trade sequencing
- Explicit cost, spread and slippage modeling
Outputvalidated strategy and risk limits
03 · Deploy
Live, measured, monitored
Go live in stages, then measure everything.
- Tick-to-trade optimization of the hot path before go-live
- FIX and broker connectivity, mirrored into each client's own MT5 account
- Staged rollouts: small size first, scaled only when live matches research
- Live monitoring of fills, latency and P&L against the backtest
Outputlive, monitored order flow
Live results feed the next research cycle. The loop never stops.
03 // LOW-LATENCY ENGINEERING
Shaving the tick-to-trade path
Tick-to-trade is the time from a market data update arriving to the order leaving the engine. We treat it as a budget: measure every stage, then remove whatever does not need to be there.
Tick-to-trade reduction
Measure every stage from packet in to order out, cut the slowest one first, then measure again.
Profiling and refactoring live strategies
Profile strategies that are already trading, find the hot spots and refactor them without changing their logic.
Multithreaded, latency-critical components
Dedicated threads for the hot path. Logging, persistence and analytics stay off it.
Lock-free and advanced concurrency
Lock-free queues, atomics and explicit memory ordering, so threads hand off data without waiting on each other.
TCP/UDP multicast and FIX
Multicast market data in, FIX order flow out, with session handling and recovery built for live markets.
Tick-to-trade pipelineIllustrative telemetry
- 01Market data inUDP multicast3.2µs
- 02Book buildnormalize1.8µs
- 03SignalC++ model6.5µs
- 04Pre-trade risklimits0.9µs
- 05Order outFIX2.4µs
p50 tick-to-trade
14.8µs
Simulated values for illustration. Not a measured benchmark.
Telemetry shown is illustrative. Real latency budgets depend on the venue, the broker and the hosting, and we measure them per deployment.
04 // OBSERVABILITY AND RESEARCH CLOUD
See the engine. Scale the research.
You cannot fix what you cannot see. Every engine we run is instrumented end to end, and research compute scales out when the backtest queue grows.
engine / live monitoring
Prometheus · GrafanaIllustrative
Illustrative monitoring dashboard showing the kind of panels we run for every engine: tick-to-trade latency percentiles, live fills compared with the backtest, order flow, and the health of each engine component.
Prometheus · Grafana
Live monitoring
Metrics and dashboards for fills, latency, positions and engine health, with alerts when live behavior drifts from research.
Investigation
Live performance issues, traced
When something looks off, we trace it from the symptom, a slow fill or unexpected slippage, to the exact stage that caused it.
AWS
Research and parallel backtesting
Backtests fan out across cloud workers, so a parameter sweep that would take weeks on one machine comes back in hours.
Tooling
Strategy, P&L, execution and latency analysis
In-house tools that compare live results with research, break down P&L and execution quality, and chart latency distributions.
Money is time · parallel backtesting
Illustrative1 workerabout 28 days
64 parallel workersabout 10.5 hours
2,000 backtests × 20 minutes each. Same work, same code: the only difference is how many workers run it at once.
05 // NON-CUSTODIAL EXECUTION
We run the technology. Never the money.
The QuantFX engine mirrors trades directly into each end client's own MT5 account at their own regulated broker. The client keeps full control of the account and the funds. We never touch them.
Licensed technology
QuantFX engine
Strategies, risk checks and execution logic, operated and monitored by QuantFX.
Orders only
Trade mirroring
Each trade is mirrored as an order instruction. No money moves through QuantFX.
Own regulated broker
Client's own MT5 account
The client opens, funds and controls the account. Deposits and withdrawals stay between the client and their broker.
No custody
We never hold, custody, manage or withdraw client funds.
Client's own broker
Accounts sit with the client's chosen regulated broker, in the client's name.
Technology, not advice
We license and operate technology. We do not give investment advice.
Client in control
The account, the funds and the decision to stay connected remain with the client.
QUANT-FX LTD is a technology and IP-licensing company, not a regulated financial firm. We are strictly non-custodial: we never hold, custody, manage, or withdraw client funds, and we do not provide investment advice. Trades are mirrored into each end client's own account at their own regulated broker. The "FX" in our name refers to asset-class focus only. Trading involves substantial risk of loss.
06 // FAQ
Technology questions, answered
How we build, test and run trading systems, and where the line sits between our technology and our clients' money.
Yes. Our quant developers research, build, backtest and deploy bespoke systematic trading strategies and trade-copy engines for companies. The same team owns the code from research through production in C++17/20, and we can operate it for you once it is live.
C++17 and C++20 for strategy and execution code, with multithreading and lock-free concurrency on the hot path. FIX protocol and TCP/UDP multicast for market data and order connectivity, MetaTrader 5 (MT5) for client accounts, Prometheus and Grafana for live monitoring, and AWS for research and parallel backtesting.
Every strategy runs through parallel backtests across instruments, parameters and market regimes, then walk-forward analysis, out-of-sample holdouts and Monte Carlo resampling, with explicit cost, spread and slippage modeling. Strategies that pass go live through staged rollouts and are monitored against their backtest.
Tick-to-trade latency is the time between a market data update arriving and the resulting order leaving the system. We reduce it by measuring every stage of that path, profiling and refactoring the hot code, keeping slow work off latency-critical threads, using lock-free handoffs between threads, and tuning TCP/UDP multicast and FIX connectivity.
No. QuantFX is strictly non-custodial. Our engine mirrors trades into each end client's own MT5 account at their own regulated broker, and the client keeps full control of the account and the funds. We never hold, custody, manage or withdraw client money, and we do not give investment advice.
Yes. We offer white-label managed quant infrastructure, or trading infrastructure as a service: we operate the strategies, engines, execution technology and monitoring on your behalf. We manage the technology only, never the money. Trades are mirrored into your clients' own accounts at their own regulated brokers.
Yes. Companies license the QuantFX engine to mirror our systematic strategies into their end clients' own MT5 accounts at their own regulated brokers. It is licensed technology: the clients keep their accounts and funds, and we run and monitor the engine.
07 // Work with us
Have a strategy to build, or an engine to run?
Tell us what you want to build. We scope the research, the C++, the test plan and the infrastructure, then run it with you or for you. Technology only, never the money.
Need quant developers on your project?Hire our quant team