The roleYou are the go-to software engineer our customers talk to when the model has to actually work in their environment and operations. You join their calls, figure out what's really broken, and ship the fix - integration code, deployment config, or a patch to our platform.
Most of the work starts with Language Translation: real-time speech-to-speech translation running on customer infrastructure. Increasingly it's Sovereign AI - Sanas AI models and solutions deployed inside a customer's own AWS account or private data center, with no data leaving their walls.
You'll work side by side with our most important customers on the deployments that matter most. One week you're sketching system architecture and shaping product direction, the next you're shipping a production integration. The problems are hard and they're yours to own end to end, from first call to live traffic.
What you'll do- Join customer calls with sales, solutions, and leadership. Diagnose technical issues live and own them through to resolution.
- Build and debug integrations between Sanas and customer systems: telephony stacks (SIP, WebRTC, media servers), contact center platforms, mobile and desktop clients, and our APIs and SDKs.
- Deploy Sanas models on customer infrastructure: their AWS, GCP or Azure accounts, Kubernetes clusters, air-gapped environments, and private data centers.
- Size and tune inference for real-time workloads - GPU and CPU footprints, latency budgets, concurrency, throughput, and cost per session.
- Profile and fix production problems across the stack: audio quality, latency spikes, model behavior, networking, scaling.
- Write the code that turns one-off customer fixes into reusable platform features, tooling, and deployment templates.
- Bring structured findings back to ML research, inference, and platform teams: what broke, why, and what we should build.
- Produce clear runbooks, reference architectures, and integration guides so the next deployment is faster than the last.
What we're looking for- A very strong software engineer first. 2+ years shipping production systems, with deep fluency in Python and at least one of Go, C++, or Rust.
- Solid grasp of ML fundamentals: how models are trained, evaluated, and served, and why they fail in the real world.
- Hands-on experience with model inference in production - serving frameworks (Triton, TensorRT, ONNX Runtime, vLLM or similar), batching, quantization, GPU utilization.
- Strong cloud and infrastructure skills: AWS in depth, Docker, Kubernetes, Terraform, or equivalent IaC, networking and security basics.
- Comfort deploying into environments you don't control - restricted networks, customer security reviews, limited access, on-prem hardware.
- Excellent debugging instincts across distributed, latency-sensitive systems.
- Clear communicator with customers and executives alike. You can explain a latency regression to a CTO and to a junior engineer.
- High ownership. You treat a customer's problem as yours until it's solved.
Nice to have- Speech and audio experience: ASR, TTS, machine translation, real-time audio pipelines, codecs.
- Prior forward deployed, solutions engineering, or customer-facing engineering role at an infrastructure or AI company.
- Telecom or contact center integrations - SIP trunks, carrier networks, CCaaS platforms.
- Sovereign or regulated deployments: air-gapped installs, data residency, healthcare or government compliance.
- Experience with GPU cluster operations, NVIDIA tooling, or CPU-optimized inference.
- Startup experience, comfort with ambiguity, and a habit of writing things down.