A small group,
held to one bar.
L² Research is an independent research pod based in the UK. We work across quantitative finance, low-level systems engineering and machine learning. Everything we ship is built to defend itself in front of industry professionals, and every member is held to that.
- Established
- 2026
- Base
- UK
- Pillars
- 00
- Intake
- Open
Why L² exists.
Most groups optimise for signalling: panels, prizes, a logo on a slide. We optimise for one thing: output a serious reader cannot dismiss.
We are not a trading club. We do not run pitch competitions. We build and we publish. Two pillars carry the work. Under research engineering, topics such as bare-metal systems, low-latency infrastructure, high-performance computing and machine learning. Under mathematical frameworks, topics such as measure-theoretic probability, stochastic calculus, game theory, optimisation and formal verification. Neither list is closed.
Every output stands on its own. The method is documented, the data is traceable, and the result can be checked by anyone who reads it. If it cannot survive a room of quants, it does not go out under our name.
Two pillars, equal weight. A stochastic calculus proof and a backtesting engine answer to the same bar: assumptions stated, method documented, result reproducible.
Membership is earned. Every member defends their output in front of the group before it ships, then defends it again in public.
We publish. Papers, source code, benchmark numbers and the failures. Work nobody can read or run is work nobody can check.
No house view, no sponsor, no mandate to please. We answer to the work, and we recruit against it, not against social credit.
A single loop, run to completion.
One workflow, deliberately narrow. Every paper, proof and system we ship has moved through the same four steps. No shortcuts, no exceptions, regardless of who is writing it.
- 01Frame
A precise question and a falsifiable claim, or a system with a target and a latency budget. Vague topics are turned back at the door.
- 02Build
Model or implementation. Assumptions written down, method chosen for reasons rather than familiarity, data provenance and source under version control.
- 03Defend
Internal review with the pillar lead. Sensitivity checks, benchmarks against a baseline, formal verification where the claim allows it, and honest limitations.
- 04Publish
Working paper, source code and benchmark numbers, in public. Failures documented as clearly as results.
- Q1 2026Founding
L² Research is formed in the UK around a small group of engineers and mathematicians.
- Q2 2026Two pillars set
Research Engineering · Mathematical Frameworks. Two tracks, one standard.
- Q3 2026First intake
Selection opens. Candidates are judged on what they have built and proved, not on credentials.
- Q4 2026First public output
Initial working papers, benchmarks and source code published under the group's name.
Each pillar has a lead who owns the standard, runs review and holds the pipeline of work published under that heading.
Two pillars. One bar.
Read the work, or join us.