About Nehme AI Labs:
Nehme AI Labs brings rigorous software engineering to AI. We audit enterprise AI architectures, cut inference costs by 40–60%, and build tools like FlashCheck that make AI systems cheaper and more reliable. We're a small, fully remote, async-first team. We hire based on proof, not pedigree.
About the Role
We're hiring a Senior Software Engineer to build and own the infrastructure behind our AI products — inference services, APIs, and internal tooling. You'll work directly with research to productionize models, make architecture calls, and set the engineering bar. This is a hands-on role: you ship, you own your work end-to-end, and you solve problems without being told how.
What You'll Do
Design, build, and operate backend services for model inference, batching, and caching
Own features end-to-end: architecture, implementation, deployment, and monitoring
Develop APIs and SDKs used by enterprise customers
Optimize inference and data pipelines for latency, throughput, and cost
Lead technical decisions on architecture and tooling, and explain the tradeoffs in writing
Mentor other engineers and raise the code quality bar through review and standards
Work with research to productionize new models
Keep production systems running and write docs that don't suck
What We're Looking For
3+ years of professional experience building and operating production software
Strong proficiency in Python, plus experience with at least one systems language (Rust, Go, C++)
You've shipped services with real users and kept them running
Solid cloud fundamentals (AWS or GCP) and containers (Docker, Kubernetes)
Sound technical judgment — you can defend a tradeoff, including the one to not build something
A track record of shipping: projects, open-source work, or products you can show us
Good async communication skills (we don't do meetings for things that should be docs)
Nice to Have
Experience with ML serving frameworks (Triton, vLLM, TGI) or LLM inference at scale
Evaluation or benchmark tooling experience
Developer tools with real adoption
Familiarity with WebAssembly or browser-based ML inference
