Senior Backend Engineer, Infrastructure

Ellipsis Health•San Francisco, CA

56d•$175,000 - $250,000•Remote

About The Position

Ellipsis Health is developing cutting-edge AI/ML products to address healthcare staffing issues and administrative burdens. The platform employs conversational AI and patented vocal biomarker technology to deliver improved healthcare experiences for clients and patients. The team is headquartered in Silicon Valley and is funded and supported by preeminent venture capital groups and healthcare institutions. The organization is open to remote candidates for this role.

Requirements

4+ years building production back-ends in modern Python.
Proven experience with real-time streaming systems—WebRTC, WebSockets, or gRPC streaming—and proficiency with asyncio, FastAPI, or similar async frameworks.
Deep understanding of concurrency, buffering, audio codecs (Opus, PCM), and distributed tracing.
Solid understanding of AWS/GCP/Azure, including container orchestration (Kubernetes/EKS/GKE), message queues (Kafka/SQS/Pub/Sub), and IaC (Terraform).
Solid grasp of relational (PostgreSQL) and in-memory (Redis) data stores; able to model and persist conversational state.
Excellent communication skills and a bias for measured, observable, and continuously deployable software.

Nice To Haves

B.S./M.S. in CS, EE, or related fields.
Familiarity with voice-agent frameworks,
Hands-on with telephony (Twilio, Telnyx), SIP, or PSTN integrations.
Experience integrating multimodal inputs (vision, text chat) into voice agents.
Familiarity with GPU inference and streaming pipelines.
Prior work in regulated industries (healthcare, finance) and comfort preparing for SOC 2 / HIPAA audits.

Responsibilities

Expand and maintain our real-time voice pipeline: design, implement, and maintain Python micro-services for conversational AI orchestration—audio capture, streaming transcription, prompt/LLM logic, synthesis, and playback.
Integrate new providers & transports: add plug-ins for emerging ASR, TTS, LLM, and memory services; wire up WebRTC, SIP, or phone endpoints; build adapters that allow hot-swapping components without downtime. Build API endpoints.
Deliver ultra-low latency (<500 ms round-trip): profile async pipelines (asyncio, FastAPI, gRPC), optimize buffering, concurrency, and back-pressure handling.
Instrument & observe every hop: emit structured traces (OpenTelemetry), metrics, and logs for each pipeline stage; define SLOs for first-token latency, end-to-end latency, and streaming reliability.
Harden for production: implement graceful retries, idempotent message passing, circuit breakers, and HIPAA-compliant security (encryption in transit, per-tenant isolation, secrets rotation).
Collaborate cross-functionally with ML, product, data engineering, and client-SDK teams to deliver features such as voice cloning, multimodal hand-offs, and domain-specific memory retrieval.