Research Engineer

This contract Research Engineer role develops algorithms for the electricity market, including solar power generation forecasting, JEPX spot and imbalance price forecasting, and optimization of trading plans that combine batteries with power-market transactions. The work is based on prepared problem definitions, datasets, and evaluation metrics, allowing the engineer to focus on improving and validating models. The scope can expand from clearly defined tasks according to progress and suitability.

Responsibilities include improving and validating solar generation models for time-series and quantile forecasting; researching and implementing forecasting models for electricity-market prices and imbalances; researching and implementing optimization algorithms for battery operation and bidding plans; backtesting with historical data; evaluating results using accuracy and revenue metrics; and reviewing papers and prior research before implementing relevant methods.

The technical environment uses Python, including pandas and polars, with scikit-learn and PyTorch as needed. The data environment includes BigQuery and GCS. Claude Code, Codex, and other AI-agent harnesses are used for code development. Required experience includes implementing machine-learning or mathematical-optimization algorithms in Python, whether through research, personal development, competitions, or other work. Candidates must also have experience developing code with AI agents and be able to read papers or prior research and independently turn them into implementations. Japanese communication must be workable for the role.

Experience with time-series forecasting, probabilistic or quantile forecasting, mathematical optimization, Kaggle or competitive programming achievements, and data analysis using BigQuery or SQL is welcome. Interest in the power or energy market is also welcomed. The role suits someone who forms and investigates hypotheses independently, iterates quickly with AI agents as tools, and can work autonomously in a remote and asynchronous setting.

The work concerns real electricity-market data and evaluates results through both accuracy and revenue, with outcomes connected to practical operations. The engineer can discuss the work directly with an engineer responsible for the full process from forecasting through bidding and operation, and direct mentoring is available. The company is a startup founded in November 2024 with a team of more than ten people. There is a path to appointment as a new graduate or experienced regular employee.

The engagement is an outsourcing contract with hourly compensation. Pay is discussed during interviews based on skills and experience. The expected workload is approximately 10 hours or more per week, with working times adjustable by the contractor. There are no prescribed working days or weekdays under the contract. The position is based in Minato, Tokyo, and remote work is available. Commuting expenses are provided, while company social insurance does not apply to this outsourcing contract. Working for three months or longer is welcomed.