▪Build and deploy custom implementations: AI Voice Assistants, Telnyx APIs (Voice, Messaging, Fax, Wireless), WebRTC
▪Run open-weight LLMs in customer environments — consume GLM, Kimi, DeepSeek, Qwen, and MiniMax through Telnyx Inference (OpenAI-compatible API; there's no model infrastructure for you to run), or deploy self-hosted stacks (Llama, Mistral, and region-specific models like CKIP-LLO, FlagAlpha, or other Cantonese/Chinese-language sovereign models) behind your own serving engine when air-gapped or sovereign-cloud requirements demand it
▪Deploy and operate LiteLLM as the model gateway in customer environments: a unified OpenAI-compatible interface across self-hosted open-weight models and hosted providers, with routing, load balancing, retries and fallbacks, rate limits, and per-team virtual keys
▪Instrument and govern LLM usage through the gateway — cost tracking and budgets, caching, logging and observability (OpenTelemetry, Langfuse, or similar), and guardrails — so customers can see and control what their AI workloads are doing
▪Wire production observability for everything you ship — metrics, logs, traces, dashboards, and alerting (e.g. Prometheus + Grafana for metrics, OpenTelemetry for traces, Graylog or ELK for logs — or the customer's existing stack) — so you and the customer's ops team both know what "healthy" looks like and what pages whom when it isn't
▪Design model routing strategies for real-time voice workloads, balancing latency, cost, and quality, with sane fallback behavior when a provider degrades
▪Make the build-vs-buy case between self-hosted open-weight models and hosted frontier APIs, and keep the customer's application code portable across both
▪Adapt models to customer domains: prompt and RAG pipelines and evaluation harnesses for English, Cantonese, and Mandarin language use cases and Hong Kong/APAC-specific regulatory environments (PDPO, data localization, Mainland China cross-border data transfer)
▪Lead POCs, pilots, and production launches from whiteboard to go-live
▪Own customer outcomes — stay engaged until the solution is live and stable
▪Collaborate directly with Product and Engineering to shape the roadmap based on field insights from Hong Kong and the broader Greater China/APAC region
▪Create clear technical documentation, runbooks, and maintainable solutions for handoff
▪Troubleshoot and resolve complex integration issues alongside customer teams
What We're Looking For
▪CS degree or equivalent experience
▪3+ years building and shipping production software — you've written code that real users depended on, you've been on-call for it, and you've debugged it when things broke. (A consulting background counts if this is also true of you.)
▪Proficiency in multiple languages: Python, Node.js/TypeScript, Go — you're more dangerous in some than others. (Telnyx Edge Compute — functions and stateful actors — is TypeScript, so that one pulls double duty.)
▪Practical understanding of what breaks in front of a model in production: provider rate limits and quotas, timeout and retry behavior, streaming, token accounting and cost attribution, and the failure modes that only show up under concurrency
▪Comfortable deploying containerized services on Kubernetes, with secrets management, config, and upgrades as part of the deployment story
▪You've wired observability and been paged because of it — Prometheus + Grafana, OpenTelemetry, Graylog, ELK, or equivalent. You know what to monitor, what to alert on, and what "healthy" looks like
▪High-concurrency experience — Kafka, message queues, or event streams at real scale. When throughput spikes you know what breaks — consumer lag, backpressure, hot partitions, rebalance stalls — and what absorbs it: partitioning, consumer groups, parallelism up to your partition count
▪You've built APIs from scratch, not just consumed them — OpenAPI spec, REST/GraphQL design, auth, rate limiting, versioning, idempotency
▪Event-driven thinking and cloud-native instincts
▪Exposure to SIP, WebRTC, or real-time voice/messaging systems
▪Self-sufficient by default — there's no engineering team behind you. You read the code and the docs, ask the customer (not your manager), and make sound engineering judgment calls on your own
▪Customer-facing engineering experience — you can run discovery with a customer's engineers and present to their executives in the same week
▪Ability to translate "it's not working" into root cause
▪Comfortable working on customer sites and in high-stakes technical conversations
▪Excellent written and verbal communication in English and Chinese (Cantonese and/or Mandarin). You can run whiteboard sessions, live troubleshooting, and escalations with customer engineering teams.
▪Based in Hong Kong, or willing to relocate — this is a hybrid role with travel across Hong Kong and the broader Greater China/APAC region
▪Legally authorized to work in Hong Kong, or eligible for sponsorship
Bonus Points For
▪Experience with AI voice assistants, STT/TTS, or LLM-based conversational systems
▪Building eval sets for a specific domain
▪Hands-on production experience with an LLM gateway — LiteLLM, Portkey, Kong AI Gateway, or an in-house OpenAI-compatible proxy — including config-driven model definitions, routing and fallback rules, and virtual key management
▪Familiarity with the open-weight model landscape and the trade-offs between regional and multilingual models, including Cantonese/Chinese-language models and Greater China sovereign AI initiatives
▪Experience with an inference serving engine behind the gateway — vLLM, SGLang, or TGI for production serving; Ollama for lightweight on-prem — and sizing GPU capacity for self-hosted models (KV-cache footprint, concurrent batch size, and context length against VRAM)
▪SQL proficiency (Postgres, MySQL, Oracle)
▪ETL and data wrangling experience
▪CI/CD pipeline design and automation
▪Background in telecom, CPaaS, or high-growth SaaS
▪Experience with sovereign or on-prem cloud deployments and Hong Kong/APAC regulatory frameworks (PDPO, Hong Kong data localization, Mainland China PIPL and CSL cross-border data transfer requirements, Office of the Government Chief Information Officer cloud guidelines, APAC cross-border data transfer frameworks)
▪Security mindset (IAM, encryption, audit logging)
#LI-RH1