Jaume Vaquer

I build the quantitative systems behind markets that clear in real time — power markets, and the trading desks that run on the same machinery.

Barcelona  /  C++  /  Python  /  AWS

00h 06h 12h 18h 24h charge dispatch
A typical Iberian day-ahead price curve: solar collapses the midday price, demand rebuilds it after sunset. The flexibility systems I build absorb energy in the trough and release it into the peak.

What I do

Strategy, optimisation and execution

Bidding and dispatch logic for day-ahead, intraday and balancing markets. Revenue optimisation against contracted capacity, deviation management, and corrections across the intraday and continuous sessions — including the solver underneath, when a commercial licence is not the right answer.

day-ahead · intraday · continuous · aFRR · MILP · SCIP

Demand-side flexibility and virtual power plants

Turning distributed assets — heat pumps, water pumps, HVAC — into dispatchable storage. Device telemetry through to market decision, including the cloud infrastructure that carries it.

demand response · VPP · aggregation · IoT telemetry · AWS

Low-latency C++ and market data infrastructure

Systems where the cost of being slow is measurable, and the market data pipelines that feed them. Five years of production C++ across defence, supercomputing and energy trading — the same constraints a trading desk works under.

C++17/20 · market microstructure · real-time pipelines · AWS

Selected work

Axpo2025 — present
Energy trading

Demand management for a 15 MW flexibility portfolio

Built the system that treats industrial and commercial devices as a virtual battery: forecasting the demand curve, deciding when to shift consumption, and correcting the position in the intraday and continuous markets. Strategy design covers contracted capacity, deviation cost and participation in secondary regulation.

Around 60 devices under management across a portfolio growing by at least 30 MW a year.

I also wrote the optimiser it runs on. Replacing the commercial solver with Geo — my own engine, built on SCIP — removed €45,000 a year in licensing while keeping the dispatch decisions the portfolio depends on.

C++ · Python · AWS · SCIP · IoT integration

Defence sector2023 — 2025
Naval systems

Naval chart rendering, rebuilt from 0.2 to 120 frames per second

The existing renderer drew licensed S-57 navigational charts at 0.2 fps — unusable for a vessel under way. I rebuilt it in C++ around the cartographic kernel to hold 120 fps in real time, a 600× improvement, and the display became something an operator could actually navigate by.

Alongside it, the Python microservices and AWS deployment pipelines behind the system, plus the operator-facing frontend.

C++ · Python · AWS · S-57 · real-time rendering

GTD2021 — 2023
Engineering

Wind farm optimiser, and firmware for supercomputing silicon

For a 10 MW Iberdrola wind farm, an optimiser that set blade pitch against live wind conditions and the expected production curve at spot prices. Because a blade can always be feathered into the wind, output could be held exactly to schedule — which took the cost of overgeneration to zero.

Separately, C firmware for server-class chips deployed in the MareNostrum supercomputer in Barcelona.

C++ · C · embedded systems · pitch control

Independent work

Polymarket microstructure

Three months of market microstructure analysis on prediction markets, and an exploration of how autonomous agents can generate and then validate trading hypotheses rather than just execute them. The same question that matters in power markets: how do you know a signal is real?

Python · market microstructure · LLM agents

Noised

A side project I run to learn the half of software that engineers usually never touch: growing and keeping a community alive. It is a niche social platform for concert reviews with around 1,500 active users, and I handle both the backend and the community side of it — which has taught me more about retention and product decisions than any backlog ever did.

backend · community · product

Nigul — a 3D graphics engine

A rendering engine written from scratch in C++. Built before language models could write code, which is rather the point: it is where the systems-level C++ came from.

C++ · OpenGL · real-time rendering

Background

Jaume Vaquer

I started at the silicon layer and moved up. Firmware and embedded control at GTD, real-time C++ for critical systems in defence, and now the market layer at Axpo — deciding what a portfolio of physical assets should do against a price curve.

That means I can follow a decision from a device on a rooftop to a bid in a market session, which is where most energy software projects lose their thread.

Electricity is the market I know deepest, but the machinery is not specific to it. Order books, microstructure, latency budgets, position and risk under hard constraints — a power desk and a financial one are running the same problems against different underlyings, which is why the work travels.

Degree in Computer Engineering, Universitat Pompeu Fabra, Barcelona. I work in English, Spanish and Catalan.

Available for contract work in energy and financial markets.