Applied AI Engineer

About Slang AI

Slang AI is redefining customer engagement through conversational AI, making every interaction seamless and efficient. Our mission is to transform the restaurant industry by providing the ultimate voice AI solution for consistently outstanding customer experiences.

At Slang AI, we believe how we build matters just as much as what we build. We foster a culture rooted in hospitality, ownership, and clarity, where every “Slangsta” feels valued, supported, and connected to the broader impact of our work in the AI-powered future of restaurants.

What Your Experience Will Be

You'll join Slang's Engineering team as our Applied AI Engineer, the person who turns foundation models into products that restaurants trust with their phones. This is not a research or model-training role. You'll work with hosted frontier models and open-weight alternatives, and our own stack to solve real product problems: choosing the right model for each job, designing prompt architectures that hold up across model versions, wiring up agents and tool calls, and proving with evaluation that every change makes the product better before it reaches a guest. When a caller asks for a table for four on Friday at 7, you're the reason the AI gets the date right, sounds like a great host, and knows when to hand off to a human.

You'll lead the work to put model calls behind clean interfaces, redesign prompts so they travel across providers, and build the evaluation and A/B testing machinery that lets us swap models with confidence instead of hope. You'll own this loop end to end: prototype, eval design, rollout, production monitoring, and the post-launch question of "is it actually working for restaurants?" You'll partner closely with backend and platform engineers, product managers, and conversation designers across our delivery pods, and you'll set the technical direction for applied AI across the team.

You'll report to our VP of Engineering, a leader with deep experience across engineering and product who cares about craft, directness, and shipping work that matters. Our stack runs on Google Cloud Platform and spans multiple languages; we pick the right tool for the job and don't require experience in any specific one. You'll join key ceremonies including daily standup at 1:00 PM ET, bi-weekly cycle planning, and weekly Show & Tell. The role is hybrid from New York City (3 days a week in our NYC office), working alongside a distributed team across the US and Brazil.

Why You Belong Here and How You Will Grow

We're AI Native: we don't just sell AI, we run on it. That means your work isn't a side experiment bolted onto a product; it's the product. Our engineering culture values pragmatic solutions, low ego, and quality as a form of respect for whoever builds on your work next. We test before we ship, we ask "why" before we commit, and we'd rather finish a few meaningful things than start many. You'll find teammates who share what they learn in Show & Tell and written design docs, and who treat a hard question from a colleague as a gift, not a threat (Humility Ability is a real value here).

You'll belong here if you get excited about the gap between a great demo and a great production system: latency budgets measured in milliseconds, evals that catch the regression nobody noticed, and a prompt change that quietly lifts booking accuracy across thousands of calls. Because we Build for Hospitality, you'll spend time listening to real calls and learning how restaurants actually run, so your technical judgment is grounded in the guest's experience.

At L5 you'll grow toward staff-level impact: setting applied AI standards that other engineers build within, influencing product strategy with what the models make newly possible, and mentoring engineers who are ramping into AI work. You'll help decide whether and when to fine-tune, self-host, or build more of our own AI capability, and you'll be the person we trust to make that call well.

What Success Looks Like

You will design and ship a provider-agnostic model interface for our voice and text AI, so we can run the same product on different models without rewriting prompts or business logic.

You will build an automated evaluation pipeline (offline eval sets, scenarios, production metrics) that every model and prompt change runs through before it ships, and that the rest of the team can use without you.

You will run model and prompt experiments in production with clear success metrics (booking accuracy, containment, transfer rate, caller hang-ups, latency, cost) and make recommendations the team can act on.

You will measurably reduce AI errors that matter most to restaurants, by moving critical facts into deterministic tools and guardrails rather than model memory.

You will own the latency and cost profile of our AI stack, and bring a clear point of view on when to use frontier APIs, faster inference providers, or open-weight models.

You will set the technical standards for prompt architecture, agent design, and AI observability, and document them so engineers across pods can build on them.

You will level up the team: mentoring engineers ramping into applied AI and bringing relevant developments in the AI tooling landscape back to Slang, with a point of view on what's worth adopting.

What You Will Bring

Required

7+ years of professional software engineering experience, including 2+ years building and operating LLM-powered applications in production (OpenAI, Gemini, Anthropic, or open-weight models).

A track record of technical leadership at the Lead / Tech Lead level: making design decisions with long-term implications, guiding implementation across a system, and mentoring senior engineers.

Deep, hands-on prompt engineering and agent design experience: structured outputs, tool/function calling, multi-step agents, and prompt designs that survive model upgrades. Familiarity with frameworks like DSPy, LangGraph, or similar agent frameworks is a plus; strong fundamentals matter more than any one framework.

Experience designing AI evaluation systems: building eval datasets, choosing between deterministic checks and LLM-as-judge, running A/B tests, and using eval or observability platforms such as Braintrust, Cekura, or similar.

Experience working across multiple model providers and abstracting them behind clean interfaces (for example with LiteLLM, OpenRouter, or your own routing layer), with sound judgment on latency, cost, and quality trade-offs.

Understanding of real-time voice AI architectures: how ASR, LLM, and TTS components fit together, and how speech-to-speech models change the picture.

Strong programming skills in at least one backend language, and comfort ramping quickly into others.

AI fluency: you actively use AI tools in your daily engineering work (coding assistants, agents, AI-assisted review and testing), with clear judgment on when to lean on them and how to validate their output.

Preferred (nice to have)

Production experience with speech-to-speech or real-time voice frameworks.

Experience with streaming ASR/TTS providers or open-source models.

Experience hosting and serving open-weight models, or evaluating inference providers on cost, l