The Role That Doesn’t Exist (But Maybe It Does)

🧠 The Role That Doesn’t Exist (But Maybe It Does)

I’ve spent years as a Technical Director. Debugging scenes. Building pipelines. Creating tools that make the creative process flow smoother. You live between the artist and the machine. You don’t make the final image, but if you do your job right, someone else can make it faster, better, and with fewer headaches.

As I’ve been looking into AI, and I see a strangely familiar landscape.

There’s an explosion of energy around prompt engineering. People are learning how to write better instructions for LLMs, shape tone, and guide outputs. It’s powerful, but focused on the surface, what you say to the model and how it responds.

At the other end, there’s the AI architect. They build infrastructure, integrate APIs, manage retrieval systems, and design agent orchestration. Heavy engineering. Full-stack thinking.

But in between those two poles, I see a gap.
And I can’t help but wonder: where’s the person who connects them?

The person who lives comfortably between instructions and infrastructure. Who bridges human creativity and machine logic. Who can shape a prompt like a storyteller, but code for days to make sure it scales.
That’s the AI TD.

---

🧩 Why the AI TD Matters

This role isn’t just theoretical. It’s already happening, quietly, in teams that don’t have a name for it yet. It’s the person:

• Building pipelines to preprocess messy inputs
• Optimizing LLM usage for latency, cost, and accuracy
• Deploying models in production environments
• Collaborating with prompt engineers to operationalize ideas
• Working with AI architects to implement modular systems
• Designing reusable prompt libraries, scoring frameworks, and test suites

It’s a role that demands the creative agility of a writer and the obsessive systems-thinking of a developer. You’re not just the bridge. You are the connective tissue.

If you’ve ever designed a scene graph tool, built a rigging interface, or scripted 200 lines just to shave five minutes off a review session, you already know the mindset. You already are this role. It just never had a title.

Until now.

---

🎛️ Enter: Prompt Playground

To explore what this AI TD role might look like, I’ve been building something called the Prompt Playground. It’s a testbed to explore how prompts behave under different conditions and across different models.

It starts simple.

• Write structured prompts in Markdown
• Run them across local (Mistral), web (ChatGPT, Claude, Gemini, etc.), and API-connected models
• Log outputs, score behavior, and refactor language like you would refactor any pipeline asset

But that’s just the surface.
Underneath, I’m designing something more powerful.

• I’m building my own vector database to manage contextual memory, custom embeddings, and structured knowledge tailored to creative workflows.
• The next step is creating a production-ready RAG system, where retrieval isn’t just a trick but a tested layer in creative pipelines, something that artists, educators, and producers can depend on.
• And when the system is solid, I’ll bring it full circle by fine-tuning a model, trained on domain-specific data that reflects how we actually work in production.

Each step is designed like a pipeline, built for iteration, built for others, and built with real-world pressure in mind.

Because if you want smarter AI in your workflow, it’s not just about prompts. It’s about the full stack behind them.

---

If you’ve ever been the person smoothing workflows behind the scenes, translating between artists and code, you already understand this mindset. You might not call yourself an AI engineer, but you’re doing the kind of work that role demands.

Maybe it’s time we gave it a name.
Or maybe the work will name itself once enough of us start doing it.

---