Principal Software Engineer — Backend & Infrastructure - Noida / Bengaluru
Level, India, Remote
What you will do
What you'll do
- Own the architecture for real-time data processing at scale. Design and evolve distributed messaging systems that handle high-throughput streaming workloads with strict latency requirements.
- Build the ML platform that lets us ship models faster. Define and execute the technical roadmap for training and serving infrastructure as our models grow in size, complexity, and inference cost.
- Scale our GPU infrastructure. Own capacity planning, scheduling, and utilization across training and inference fleets — deciding what runs where, how we handle burst demand, and how we keep spend tied to actual throughput rather than idle reservations.
- Scale inference. Drive down latency and cost per request as model complexity and traffic grow: batching and routing strategies, quantization and compilation, autoscaling, and caching — without degrading output quality.
- Make reliability a property of the system, not a heroic effort. Drive uptime, observability, and incident response for serving systems that enterprise customers depend on in production.
- Turn ambiguous business problems into executable technical plans. Partner with Product and GTM to scope large cross-functional initiatives, then break them into work other teams can run with.
- Multiply the team. Lead design reviews, mentor senior engineers, and shape the engineering practices that outlast any single project.
- Bring the outside in. Evaluate emerging tools and techniques with judgment — adopt what earns its complexity, skip what doesn't.
- Security and Compliance – Comply with Level AI’s information security, privacy, data protection, and acceptable-use policies.Protect company, customer, and confidential information; use only approved systems and access information strictly based on business need. Complete required security and privacy training and promptly report any suspected security incident, data exposure, or policy violation.
What success looks like in your first year
- 90 days: You've mapped our critical paths and failure modes, shipped a meaningful improvement to a serving or pipeline bottleneck, and earned trust across both sites.
- 6 months: You own a published technical roadmap for [messaging / ML infra], with at least one major migration or redesign underway. GPU utilization and inference cost per request are measured, and trending the right way.
- 12 months: Our infrastructure scales predictably with customer growth, inference cost stays flat or falls as volume rises, on-call load is down, and other engineers are making better architectural decisions because of standards you established.
What you need
- [9-12] years building backend and infrastructure systems, with a track record of owning architecture and design at scale — not just implementing it.