01
Market data pipelines
Venue adapters, normalisation to a single schema, capture with hardware timestamps where the feed supports it, and a storage layout a five-year query can survive. We read the ITCH spec so your researchers never have to.
Market data · Execution · Research tooling
We build the layer between the venue and your strategy.
Venue adapters, normalisation, capture and replay, FIX connectivity and the research harness — built inside systematic trading firms. Your engineers work on alpha instead of parsing another exchange's binary quirks.
# one schema for every venue — the adapter # layer we build and hand over to you from marketdata import Book book = Book.replay( venues=["XNAS", "XLON", "XETR"], date ="2026-08-14", schema="mbp-10", ) for tick in book: tick.ts_event # ns, exchange clock tick.ts_recv # ns, our capture tick.levels[0] # bid/ask, size
OneSchema, every venue
nsTimestamp resolution
SameCode path, research and prod
YoursSource, tests and docs
What we build
The unglamorous half of a trading system.
None of this is where your edge lives — which is exactly why it shouldn't be consuming your quant engineers. It still has to be correct to the nanosecond, and it still decides whether you trade on Monday.
02
Execution & FIX connectivity
FIX 4.2 through 5.0 SP2, session management that survives a venue restart, order routing, drop copy, and the certification runs with each venue. Tedious, exacting, and the thing that breaks at 09:30 if it's done casually.
03
Research tooling & replay
A backtest harness that reads the same bytes as production, deterministic replay at wall-clock or accelerated speed, and point-in-time symbology so a result from last year still reproduces this year.
04
Run it with you
Session monitoring across your trading calendar, gap detection, venue change management and an on-call rota — or a documented handover to your own team once it's stable.
Scope
A short list of things we will not do.
We write software. We don't run your cloud account, rack your hardware, or have opinions about your alpha. Being narrow is the point — it's why the work is fast and why we can be held to a spec.
Not us
- Cloud platform and networking
- Colocation, cages and hardware
- FPGA feed handlers
- Your strategy and signal research
- Regulatory sign-off
Us
- Venue adapters and normalisation
- Capture, storage and replay
- FIX and execution connectivity
- Research tooling and backtest harness
- Production support in trading hours
Engagement
Scoped, delivered in your repository, handed over.
Scope
A short paid discovery: which venues, which schemas, what your researchers actually need day to day. You get a written specification and an estimate, and you keep both regardless.
Build
Delivered in your repository with tests, documentation and a conformance run against each venue. Typically six to twelve weeks, depending mostly on how many venues you need.
Run
Optional support across your trading calendar, or a handover with runbooks. Everything we write is yours — source, tests and documentation, with no runtime dependency on us.
Get in touch
Tell us which venues.
A short note about the venues, schemas and volumes you need, and what your team is spending time on today. You will get an engineer on the first call.