Optimisation at the scale your strategy needs.
A quantum-inspired optimisation engine, built for the problems portfolio managers actually face.
Mythiq Labs turns portfolio construction, allocation, and execution problems that are too large for conventional optimisers into portfolio insights you can trade on. Built by people who understand the mathematics and the markets.
Your best portfolio is trapped in a problem too big to search.
Every allocation, rebalance, and execution schedule is a choice between more combinations than any optimiser can check one by one. As time passes and as your universe grows, the gap between a workable portfolio and the perfect one widens, and that gap is fund performance you're leaving on the table.
Size breaks the tools
Past a few hundred names and a handful of constraints, the optimiser you have either times out or quietly loosens the problem to finish.
Speed can hide quality
A fast answer tells you the clock stopped, not that the portfolio is any good.
Generic tools don't speak finance
An optimiser is not a strategy. Turning your mandate, limits and risk model into a problem it can solve well is the hard part.
We turn your hardest optimisation problems into portfolios you can trade.
Mythiq Labs applies large-scale, quantum-inspired optimisation to the problems portfolio managers face every week: building portfolios across thousands of names, allocating across mandates with real limits, and scheduling the trades that get you there. The engine searches a space of combinations far beyond the reach of conventional optimisers, and returns an answer you can trade before the market moves on.
The engine is half the work. The difference between a portfolio you can run and one you throw away is how faithfully your problem is translated for it: your limits, your trade-offs, and the shape of your mandate. That translation is our craft, and it is where we spend most of our effort.
Quantum-inspired computing, in production today.
Our engine is an Ising machine. It takes an optimisation problem, expresses it as an energy function over binary, integer and continuous variables, and samples the low-energy states of that function. It generalises the Hamiltonian a quantum annealer minimises, realised through massively parallel probabilistic sampling on GPUs rather than qubits. It runs on hardware and in the cloud, using technology that is actually available today, rather than on specialised quantum hardware that is still years from practical commercial use.
Your covariance matrix is most of our Q matrix.
A mean-variance objective already has the shape a QUBO wants: a grid of pairwise weights. The covariance supplies the off-diagonal terms, expected returns and penalties load the diagonal, and every constraint you add reshapes the grid. Encoding all of that faithfully, so that the lowest-energy state is the portfolio you actually want, is where the edge is won.
covariance structure → QUBO weights
annealing
Rooted in the physics of the Ising model
Every problem becomes an energy landscape, and the answer is its ground state. We search it the way statistical physics does, sampling states in proportion to their Boltzmann probability under a cooling schedule, with large blocks of variables updated in parallel rather than one at a time. The same machinery has solved a single problem of 112,000 binary variables.
Output you can trade, not a report in development
The answer comes back as target weights and a trade list, in a form an order or portfolio management system can act on directly. Not a spreadsheet someone needs a week to interpret while the market moves and the answer goes stale.
Constraints inside the energy function
Binary, integer and continuous variables, with linear and quadratic constraints expressed directly in the energy function, so budget, name count and exposure limits shape the search from the first sweep rather than being checked at the end.
Same physics, ordinary silicon
Runs on GPUs and in the cloud with no cryogenics, no dilution refrigerators and no waiting for qubit counts to catch up. The mathematics is shared with quantum annealing; the infrastructure is what you already have.
Foundational research: Patel, Canoza and Salahuddin, Nature Electronics 5, 92 (2022); Patel et al., arXiv:2008.04436, benchmarked against the D-Wave 2000Q and the optical Coherent Ising Machine. Benchmarks on your own problem are available on request.
Where it earns its place on the desk.
Portfolio construction at scale
Build the portfolio you want across thousands of names, with sector, country, liquidity, turnover and position limits binding at once, plus the ones that break ordinary optimisers: a cap on the number of names, minimum holdings and round lots. And get it back in time to trade it.
SMA and mandate allocation
Optimise many separately managed accounts and mandates at once rather than one at a time: each with its own limits and exclusions, all drawing on the same liquidity, and a set of allocations that respects every one of them rather than a compromise between them.
Robust optimisation
Build the portfolio that holds up across several views of risk at once, not the one that is only optimal under a single covariance estimate, with your name count and lot size limits still binding. Ordinary optimisers make you choose between the two.
Execution scheduling
Sequence and size the trades so a large rebalance lands with less market impact and less cost, without the schedule breaching risk or cash constraints along the way.
Built for finance. Governed for trust. Measured honestly.
Built for finance
We don't hand you an optimiser and leave you to it. Your mandate, limits, universe and risk model go in; a tradeable portfolio comes out. The translation in between is our job, not yours.
Trust and jurisdiction
Your positions and your strategy stay within a governance regime you are comfortable with: hosted where you can see it, governed under Singapore law, with no ambiguity about who has access, or who is answerable.
Honest feedback
We won't quote you a benchmark we haven't run on your problem. Bring it, and we run it against the process you use today, on your data, under conditions we agree together. You keep the results either way.
We came to optimisation from the buy-side.
Mythiq Labs was founded by a team that has built trading technology inside hedge funds, led financial-industry standards work, and shipped systems that funds run every day.
Mythiq Labs is a deep-tech company headquartered in Singapore, one of the world's most trusted financial and technology jurisdictions, bringing quantum-inspired optimisation from research into production for the financial industry. Serious technology, from a counterparty you can diligence.
One engine, many problems.
The maths doesn't care what the variables represent. The same engine that optimises portfolios extends naturally to other industries built on hard combinatorial problems.
Tell us the problem your current tools cannot solve.
We will show you what large-scale, quantum-inspired optimisation does with it.