About Loon
Loon is on a mission to connect people everywhere by inventing and integrating audacious technology. Loon is built on a sophisticated system of products and services that combine advancements in materials science, atmospheric modeling, machine learning, and communications systems. These platforms enable Loon to harness the stratosphere, transforming global connectivity and creating endless possibilities for new applications by bridging ground, sky and space.
Overview
Loon’s network orchestration stack is a Temporospatial Software-Defined Network (TS-SDN) for networks that span land, sea, air, and space. It was designed from the ground-up to support multiple aerospace networking projects, which until relatively recently included non-geostationary satellite constellations, in addition to the balloon-based production network that it orchestrates today at Loon.
Large enterprises and startups are racing to build their own non-geostationary constellations of satellites and high-altitude platforms to provide affordable, broadband Internet service to the billions of people that still lack access to it. But no single solution can solve the problem; just as population densities on Earth are non-uniform and span roughly five orders of magnitude, the tradeoff between coverage and capacity-density will require five altitude layers (towers, stratospheric platforms, low-earth orbiting, mid-earth-orbiting, and geostationary orbiting satellites).
Enabling the stratospheric layer and the ability for high-altitude platforms to internetwork with the layers above and below is of key strategic interest to Loon. As such, Loon is externalizing its network orchestration stack and transforming it into a multi-tenant production service to enable the internetworking and coordination of the world’s aerospace networks.
Loon is seeking an experienced Software Engineer with graph theory, operations research, and machine learning experience to lead development of our next-generation network optimization engine for satellites and high-altitude platforms. This engine will determine the schedule of beam handovers, radio frequency channel assignments, and network paths that should be used, over time, by a heterogeneous network of ground stations, satellites, balloons, solar gliders, maritime vessels, commercial airlines, and end user devices on Earth. It must also minimize network disruption and packet loss as the network continuously evolves -- and rapidly repair the network in the face of any unpredicted link failures.
Responsibilities
- Investigate and down-select from a number of possible strategies for the next-generation solver.
- Align stakeholders and build confidence in your proposed strategy.
- Lead a small team of 2-3 software engineers (inclusive) to train and turn-up a production serving model within ~18-24 months.
Minimum Qualifications
- Strong understanding of reinforcement learning and contextual bandits problems.
- Production experience training and serving traffic using ML models.
- Experience designing large scale systems and implementing core components.
- Fluency in one of {C++, Go, Java}, with interest in developing proficiency in all three.
Preferred Qualifications
- Experience with or interest in aerospace systems including communications satellites.
- Experience with or interest in networking, including wireless communications and software defined networks,