Forward Deployed Engineer, UAE

Role Overview

Forward Deployed Engineer, UAE at Telnyx. Based in Dubai, this is a hybrid role with travel across the UAE and broader GCC region.

Company Overview

Telnyx is an industry leader building the future of global connectivity. The company operates a private, global, multi-cloud IP network and delivers hyperlocal edge technology through intuitive APIs. Telnyx is financially stable and profitable, investing in pioneering technologies and fostering continuous learning and growth for its team.

Role Purpose

You will embed directly with enterprise customers in the UAE to architect, build, and deploy production systems on Telnyx's global network—voice, messaging, AI, and wireless. Working alongside a dedicated Enterprise Account Executive, you will own technical discovery, architecture design, proof-of-concept development, and production go-live. You will lead with outcome conversations, not product menus, helping customers build real solutions that work at scale.

Key Responsibilities

Customer Engagement & Discovery

• Embed with enterprise customers to understand their communications workflows, AI use cases, and integration challenges firsthand
• Lead proof-of-concept, pilot, and production launches from whiteboard to go-live
• Own customer outcomes—stay engaged until the solution is live and stable
• Run discovery with customer engineering teams and present to their executives
• Translate customer problems into technical solutions and troubleshoot complex integration issues alongside customer teams
• Collaborate directly with Product and Engineering to shape the roadmap based on field insights from the UAE and broader GCC

Architecture & Implementation

• Design custom implementations including AI Voice Assistants, Telnyx APIs (Voice, Messaging, Fax, Wireless), and WebRTC
• Build and deploy containerized services on Kubernetes with secrets management, config, and upgrades as part of the deployment story
• Design model routing strategies for real-time voice workloads, balancing latency, cost, and quality with sane fallback behavior
• Make the build-vs-buy case between self-hosted open-weight models and hosted frontier APIs
• Keep customer application code portable across both self-hosted and hosted approaches
• Build APIs from scratch—OpenAPI spec, REST/GraphQL design, auth, rate limiting, versioning, and idempotency

AI & LLM Operations

• Run open-weight LLMs in customer environments—consume GLM, Kimi, DeepSeek, Qwen, and MiniMax through Telnyx Inference (OpenAI-compatible API), or deploy self-hosted stacks (Llama, Mistral, and region-specific models like Jais, Fanar, or other Arabic-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, 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
• Adapt models to customer domains: prompt and RAG pipelines and evaluation harnesses for Arabic and English language use cases and UAE/GCC-specific regulatory environments (UAE PDPL, Dubai data sovereignty, cloud localization requirements)

Observability & Operations

• 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
• Instrument observability and be on-call for it

Documentation & Knowledge Transfer

• Create clear technical documentation, runbooks, and maintainable solutions for handoff

Qualifications & Experience

• CS degree or equivalent experience
• 3+ years building and shipping production software—you have written code that real users depended on, been on-call for it, and debugged it when things broke. (A consulting background counts if this is also true of you.)
• Exposure to SIP, WebRTC, or real-time voice/messaging systems (bonus)
• Experience with AI voice assistants, STT/TTS, or LLM-based conversational systems (bonus)
• 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 (bonus)
• 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) (bonus)
• Background in telecom, CPaaS, or high-growth SaaS (bonus)
• Experience with sovereign or on-prem cloud deployments and UAE/GCC regulatory frameworks (UAE PDPL, Dubai Data Law, TDRA cloud regulations, GCC data sovereignty requirements, free zone IT environments like DIFC, Dubai Internet City) (bonus)

Skills & Competencies

• Proficiency in multiple languages: Python, Node.js/TypeScript, Go—you are more dangerous in some than others. (Telnyx Edge Compute—functions and stateful actors—is TypeScript, so that one pulls double duty.)
• Comfortable deploying containerized services on Kubernetes, with secrets management, config, and upgrades as part of the deployment story
• 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
• 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)
• Event-driven thinking and cloud-native instincts
• Self-sufficient by default—there is 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
• 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 Arabic. You can run whiteboard sessions, live troubleshooting, and escalations with customer engineering teams.
• SQL proficiency (Postgres, MySQL, Oracle) (bonus)
• Building eval sets for a specific domain (bonus)
• Familiarity with the open-weight model landscape and the trade-offs between regional and multilingual models, including Arabic-language models (Jais, ALLaM, Fanar) and sovereign AI initiatives in the GCC (bonus)
• ETL and data wrangling experience (bonus)
• CI/CD pipeline design and automation (bonus)
• Security mindset (IAM, encryption, audit logging) (bonus)

Additional Information

• Based in Dubai, or willing to relocate—this is a hybrid role with travel across the UAE and broader GCC
• Legally authorized to work in the UAE, or eligible for sponsorship