ML Researcher (Part-time)

Loka Labs; United States; Remote; USD 60,000 - 90,000; Part time

Benefits offered by this job

Remote-friendly

Part-time with weekly cadence

Work with Joseph Sifakis

Job summary

Loka Labs is building an Economic World Model and seeks a Founding Scientist to turn its architecture into a runnable time-series and behavioral simulator. You will design and implement time-series models, extend the causal framework, and create an end-to-end pipeline to backtest NGO/public funding scenarios.

This is a part-time, remote-friendly role with a dependable weekly cadence. You’ll collaborate with Joseph Sifakis and the founding team, translate research ideas into production-ready

Qualifications

• PhD (in progress or completed) in ML, Statistics, Computational Economics, Complex Systems, or related field.
• Strong capabilities in time-series modeling and/or multi-agent/behavioral modeling.
• Able to independently turn research ideas into runnable, backtestable code (Python / PyTorch, etc.).
• Reliable weekly time investment and on-time delivery; part-time is fine, but cadence must be dependable.

Responsibilities

• Design and implement time-series models (evolution of economic/social indicators) and behavioral models.
• Integrate models into the overall world-model framework – causal layer, state representation, interventions & counterfactuals.
• Get an end-to-end pipeline running in a bounded scenario and backtest it against historical data.
• Deliver on a stable weekly cadence and participate in reviews.

Skills

Time-series modeling

Multi-agent modeling

Python

PyTorch

Backtesting

Education

PhD (in progress or completed)

Tools

Python

PyTorch

Job description

Run the world, before you run it.

Every complex system gets tested before deployment. Aircraft. Semiconductors. Drugs. Self-driving cars. Except one - the economy. The architecture is already laid out in a paper; now we need to turn it into a core model that runs and backtests.

Loka is building an Economic World Model - a differentiable simulator of economic reality that "pre-runs" major economic and public-policy decisions before they happen. The core is not yet another forecasting model, but an orchestration of causal reasoning, multi-agent dynamics, and time-series & behavioral modeling that outputs a set of possible futures plus actionable levers. Our Founding Scientist is Joseph Sifakis, recipient of the Turing Award and pioneer of formal verification. First deployment scenario: public funding / aid allocation decisions.

The work

Turn the architecture into a model that runs.

• Design and implement the time-series models (evolution of economic / social indicators) and behavioral models (multi-agent responses to interventions).
• Integrate these models into the overall world-model framework - causal layer, state representation, interventions & counterfactuals.
• Get an end-to-end pipeline running in a bounded scenario (NGO / public funding allocation) and backtest it against historical data.
• Deliver on a stable weekly cadence and participate in key reviews.

Requirements

• PhD (in progress or completed) in ML, Statistics, Computational Economics, Complex Systems, or a related field.
• Strong capabilities in time-series modeling and/or multi-agent / behavioral modeling.
• Able to independently turn research ideas into runnable, backtestable code (Python / PyTorch, etc.).
• Critical: a reliable weekly time investment and on-time delivery. Part-time is fine, but the delivery rhythm must be dependable - outputs on schedule, presence at required reviews. We are results-oriented, but we need predictable output.

Nice to have

• Any of: graph neural networks; causal inference (SCM / do-calculus); agent-based modeling.
• Background in modeling economic, financial, policy, or social systems.
• Relevant publications or demonstrable modeling projects.

What you'll get

• Work alongside Turing Award laureate Joseph Sifakis and the founding team on a direction almost no one has explored.
• Help define the core model of a new category.

Logistics

Commitment

Part-time, with a dependable weekly cadence.

Location

Remote-friendly.

Compensation

Work with

Joseph Sifakis (Turing Award) and the founding team.