Senior Backend Engineer: AI Call System
Budget: $11 – $22 USD
I’m building a telecom-grade, AI-powered outbound calling platform that must comfortably push 1–2 million calls every single day. The core of the product is a real-time, high-throughput distributed service that stitches together low-latency audio streaming, self-hosted LLM inference, and iron-clad fault-tolerance. No third-party AI APIs will be involved—every model runs on our own infrastructure.
Where we stand
• Architecture: event-driven micro-services backed by async message buses, ready to fan out thousands of concurrent call flows.
• Scale targets: sub-250 ms P99 end-to-end latency, five-nines availability, horizontal elasticity across several regions.
• Comms stack: I’m still evaluating whether VoIP, SIP, WebRTC—or a blend—will best serve us, so I need someone comfortable across all three.
• Language: the surrounding platform is written mostly in Python (FastAPI, asyncio), but I’m open to bringing in Go or Java micro-services where raw throughput demands it.
• Tooling in play: Kubernetes, gRPC, Kafka (or NATS), Prometheus/Grafana, Terraform, and GPU-backed inference servers.
What you’ll drive
1. Design and harden the distributed call-processing pipeline—shaping message topics, retry semantics, circuit-breakers, and observability hooks.
2. Integrate our on-prem LLM stack so every call can stream requests / responses in real time without external dependencies.
3. Benchmark, profile, and surgically optimise for tail latency and peak traffic days.
4. Extend the codebase with clean, well-tested modules and keep CI/CD green.
5. Coach junior devs or step into a principal role as we scale, depending on your preference.
Engagement & compensation
This can start as a scoped contract or roll straight into a longer engagement if we click. Base pay begins at $7 000, with performance-based increases and equity discussions on the table for an ongoing role.
How to apply
Drop me a brief note describing the largest real-time or distributed system you’ve built or operated—numbers, challenges, and your personal contribution. Links to public repos, tech talks, or write-ups are welcome.
Looking forward to building something game-changing together.
Where we stand
• Architecture: event-driven micro-services backed by async message buses, ready to fan out thousands of concurrent call flows.
• Scale targets: sub-250 ms P99 end-to-end latency, five-nines availability, horizontal elasticity across several regions.
• Comms stack: I’m still evaluating whether VoIP, SIP, WebRTC—or a blend—will best serve us, so I need someone comfortable across all three.
• Language: the surrounding platform is written mostly in Python (FastAPI, asyncio), but I’m open to bringing in Go or Java micro-services where raw throughput demands it.
• Tooling in play: Kubernetes, gRPC, Kafka (or NATS), Prometheus/Grafana, Terraform, and GPU-backed inference servers.
What you’ll drive
1. Design and harden the distributed call-processing pipeline—shaping message topics, retry semantics, circuit-breakers, and observability hooks.
2. Integrate our on-prem LLM stack so every call can stream requests / responses in real time without external dependencies.
3. Benchmark, profile, and surgically optimise for tail latency and peak traffic days.
4. Extend the codebase with clean, well-tested modules and keep CI/CD green.
5. Coach junior devs or step into a principal role as we scale, depending on your preference.
Engagement & compensation
This can start as a scoped contract or roll straight into a longer engagement if we click. Base pay begins at $7 000, with performance-based increases and equity discussions on the table for an ongoing role.
How to apply
Drop me a brief note describing the largest real-time or distributed system you’ve built or operated—numbers, challenges, and your personal contribution. Links to public repos, tech talks, or write-ups are welcome.
Looking forward to building something game-changing together.