No greed. No panic. No rest.
SleeplessNode is an AI quant trading company founded in 2024, with a core team from Stanford University and top Wall Street high-frequency trading firms. We make quantitative trading, once reserved for institutions, available to every investor.
In trading, nothing costs more than emotion.
Fear of missing out leads to buying high. Fear of loss leads to selling low. Most trading losses begin with one of the two.
Crypto markets never close. The sharpest moves often come when no one is watching.
SleeplessNode is made up of tens of thousands of AIs. Together they monitor the world’s major crypto exchanges and act on data alone, entering when opportunity appears and exiting when risk does.
Choose a strategy or a grid. SleeplessNode takes care of the rest.
About us
While at Stanford, SleeplessNode’s founding team found that the contest inside a high-frequency order book and a robot planning its path through a hostile environment are the same class of stochastic optimal control problem. The company grew from that finding.
The team brought the two fields together. ICME contributed stochastic control, Markov decision processes, and optimal execution. SAIL and its Autonomous Systems Lab contributed multi-agent reinforcement learning and adversarial defense. The company was founded in early 2024, with its first racks in NY4, New York. Today its systems run in New York, London, Tokyo, and Frankfurt.
SleeplessNode is focused on crypto. The market never closes, moves sharply, and is full of spoofing, which makes it the hardest test for adversarial models. SleeplessNode is the first quant fund in the world to deploy adversarial defense fully in live trading.
Team
Our core team brings together academic researchers and engineers from top proprietary trading firms.
Dr. Alexander Lin
Co-founder and Chief Executive Officer
PhD in applied mathematics from Stanford ICME and former core algorithm researcher at a top Wall Street high-frequency trading firm. His doctoral research used GPU clusters and deep state space networks to model market microstructure data at scale and generate counterfactual scenarios.
Dr. Elena Rostova
Co-founder and Chief Technology Officer
From Stanford’s AI Lab (SAIL) and Autonomous Systems Lab (ASL), where she led multi-agent reinforcement learning projects for Mars surface rovers and deep-sea autonomous vehicles. She brought adversarial machine learning into SleeplessNode’s execution engine.
Marcus V. Thorne
Chief HFT Architect
Former senior microstructure engineer at CME Group. He leads kernel bypass engineering and CME MDP 3.0 multicast optimization, letting the system communicate directly with exchange matching engines at nanosecond precision.
Funding
SleeplessNode has completed three funding rounds.
- Round
- Seed
- Date
- Mai 2024
- Amount
- 22,0 Mio. $
- Post-money valuation
- 110,0 Mio. $
- Use of funds
- Deployed the first racks in NY4, New York, and rented thousands of H100 GPUs to pretrain the order book foundation model.
- Round
- Series A
- Date
- Februar 2025
- Amount
- 115,0 Mio. $
- Post-money valuation
- 550,0 Mio. $
- Use of funds
- Expanded the multi-agent RL execution engine, connected to market data and trading interfaces at major crypto exchanges, and completed validation in spot and perpetual futures.
- Round
- Series B
- Date
- März 2026
- Amount
- 350,0 Mio. $
- Post-money valuation
- 3,2 Mrd. $
- Use of funds
- Expanded into LD4 London, TY3 Tokyo, and FR2 Frankfurt, and deployed nanosecond-scale FPGA adversarial defense across the system.
Global presence
Four nodes across American, European, and Asian hours, following the crypto market around the clock.
Equinix NY4
Secaucus, New Jersey
Our North American node, covering crypto trading during American hours.
Equinix LD4
Slough, London
Our European node, covering crypto spot and derivatives during European hours.
Equinix TY3
Tokyo, Japan
Our Asia-Pacific node, covering crypto trading during Asian hours.
Equinix FR2
Frankfurt, Germany
Our continental European node, covering European hours alongside London.
Trades are executed on major exchanges including Binance, OKX, Bybit and Gate.
Our approach
SleeplessNode reads the order book with the LOBERT architecture, treating each order message as a token and keeping its price, size, and time intact. DiffLOB diffusion models generate extreme markets from set levels of volatility and order imbalance, so strategies can rehearse before going live. The VPIN metric tracks order flow toxicity in real time, and passive orders are withdrawn as soon as it rises. Multi-agent reinforcement learning decides how each order is placed, and a low-latency system sends it.
Reads the order book
Every order placed, canceled, or filled informs the decision.
Rehearses extreme markets
Generative models simulate markets that haven’t happened yet, so strategies have faced the extremes before going live.
Sees through spoofing
Fake orders and network delays are built into training, so the AI learns to recognize real intent.
Executes in stages
Large orders are split and filled gradually to minimize price impact.
Inside 2.5 microseconds
From the latest order book snapshot to a completed inference and an order on the wire.
In-building fiber
Under 500 ns
Dark fiber cross-connects inside the data center, about 5 ns per meter.
Market data intake
50 to 150 ns
FPGA network cards strip UDP/IP headers in hardware and write data straight to memory with zero copy.
Decision
200 to 800 ns
User-space C++ with zero runtime allocation and lock-free queues. Some lightweight networks run directly in FPGA logic.
Pre-trade risk checks
20 to 60 ns
Order size, credit limit, and wash trade checks run in parallel FPGA circuits.
Order dispatch
100 to 200 ns
Order messages go straight to the exchange gateway through kernel bypass (EF_VI, DPDK).
End to end
About 2.5 µs
Contact
For any question, write to us. We will reply as soon as we can.
- Product and account support
- [email protected]
- Institutions
- [email protected]
- Press
- [email protected]
Join us
We are looking for people who do careful work. Research has to survive live trading, systems have to run around the clock, and support answers have to make sense the first time. These are the roles open now. To apply, send your CV and one piece of work you are proud of to [email protected], with the role in the subject line.
Quantitative Researcher
Study price and order-book behaviour in crypto markets and turn ideas into strategies that can be backtested and run live. Strong statistics and Python required.
Trading Systems Engineer
Own order routing, risk checks and exchange connections, where speed matters and mistakes are expensive. C++ or Rust. Trading systems experience preferred.
Risk and Compliance Analyst
Review identity checks and withdrawals, answer Travel Rule requests from exchanges, and keep track of regulatory changes in each region. Compliance experience at an exchange or payments company preferred.
Multilingual Support
Answer users by email and Telegram. Fluent Chinese and English, Japanese or Korean a plus. Shift work required.
Transparency and risk
These principles apply to every SleeplessNode product.
- Every figure has a source
- Each return, drawdown, and win rate states its time window and number of data points.
- Verification first
- Identity verification is required before withdrawal. Until then, the account is view only.
- The past is not a promise
- Historical performance reflects the past only. SleeplessNode does not guarantee any return.
Tens of thousands of AIs. Ready when you are.