Los Angeles, CA | Full-time

About Fluid Concepts

Fluid Concepts is a data research lab building the data and methods needed for frontier multimodal, human-centric, and interactive AI.

We design and produce research-grade datasets for capabilities that are difficult to capture synthetically: natural conversation, social interaction, multimodal behavior, tool use, embodied interaction, and other forms of complex human behavior.

Our work sits at the intersection of AI research, experimental design, and creative audio/video production.

The Role

As the Head of Design Implementation your role sits between research design and production execution - establishing the environments, protocols, and systems through which our datasets are collected.

Our research team defines what a dataset should accomplish and our production team executes collections at scale. You are the fundamental bridge between these two functions: turning a research objective into a process that brings out the human in humans.

This is a highly hands-on role. You’ll prototype new collection formats, evaluate recording environments, design participant experiences, write elicitation materials, work directly with talent, run pilots, and turn what works into scalable systems.

What You’ll Do

1. Design collection environments

Design the physical and technical environments in which our data is created.

This might include:

• Finding and evaluating studios, offices, homes, or other real-world recording locations

• Designing room layouts, participant positioning, and interaction geometry

• Defining microphone, wearable, video, and other capture setups

• Understanding the impact of acoustics, reverberation, environmental noise, isolation, movement, and room geometry

• Prototyping novel collection environments with engineering and production

This part of the work engages the question - how do we balancerecording quality with ecological realism and participant behavior?

2. Design behavioral elicitation and performance

Translate research goals into activities that produce the behaviors we want to capture.

You might:

• Write scenarios, prompts, role briefs, and participant instructions across both scripted, improvised and naturalistic interactions.

• Create activities and constraints that elicit specific conversational, social, or expressive behaviors

• Define casting, audition, and participant-selection approaches

• Direct talent and provide performance feedback

• Iterate on protocols when interactions feel artificial or fail to produce the intended behavior

This part of the work engages the question - How do we create the conditions in which the behavior we care about emerges naturally?

3. Define quality and build scalable systems

Determine what a successful collection looks like and turn creative judgment into processes others can execute consistently.

You’ll:

• Develop collection guidelines, quality rubrics, and gold examples

• Review pilots and collections and diagnose failure modes

• Build calibration and QA processes for producers, directors, and collection teams

• Document successful protocols and turn them into reusable playbooks

• Design approaches that can scale across hundreds of participants, many sessions, and multiple locations

This part of the work engages the question - how can we translate our craft into something scalable?

What This Might Look Like Day to Day

Imagine we are building an egocentric, wearable-based multi-party dataset for diarization, ASR, and turn-taking dynamics.

You might start by working with our research team to translate those goals into a collection design: what kinds of interactions we need, who should participate, what activities will generate realistic overlap and speaker changes, and what our recording setup needs to capture.

Next, you might identify and evaluate the right physical space. Rather than simply finding an available studio or office, you would assess whether its acoustic and environmental characteristics - reverberation, background noise, isolation, geometry, nearby activity - will produce the data we want.

You might then run a pilot with several participants, modifying the room setup, prompts, participant roles, or recording configuration as you observe what works.

Later, you might review the resulting recordings, revise casting or director guidance, identify strong and weak examples, and turn what worked into a collection protocol that can be repeated hundreds of times.

The work moves between design, site (environment), prototyping, observation, and iteration.

How You’ll Work With the Team

This role works closely across Fluid:

• Research: Translate scientific goals into practical collection methods and contribute to high-level dataset design.

• Engineering: Define capture requirements and prototype recording environments, sensors, interfaces, and tooling.

• Operations & Production: Build methods that producers and collection teams can execute reliably at scale.

You own how a collection is designed to work; the broader team partners with you to make it scientifically useful, technically robust, and operationally scalable.

Who We’re Looking For

We’re looking for someone who combines technical audio and/or video expertise, strong performance and behavioral instincts, and an ability to build scalable systems.

1. Sound Design & Capture

You may have a background in audio or recording engineering, sound design, voice design, audio direction, or creative audio.

You should be comfortable with areas such as:

• Microphone selection and placement, including multi-channel and wearable capture

• Room acoustics, reverberation, noise, isolation, and environmental recording

• Evaluating studios and real-world spaces for recording

• Understanding how participant position, movement, and room geometry affect captured speech

• Critical listening and diagnosing problems in recorded speech, environments, and signal chains

2. Behavioral Elicitation & Performance

You should also have strong instincts for the human side of collection:

• Cultivating authentic performance from performers and non-performers alike

• Writing prompts, scenarios, role briefs, and participant instructions

• Designing scripted, semi-scripted, and improvised interactions

• Eliciting conversational, social, and expressive behaviors

• Casting, auditions, and participant selection

• Directing talent and evaluating naturalness, expressivity, and interaction quality

3. Scalable Systems

You should be able to turn successful experiments into repeatable systems:

• Collection protocols and playbooks

• QA frameworks and calibration methods

• Gold examples and standards

• Workflows that producers and collection teams can execute consistently across people and locations

We’re particularly excited by people who can move fluidly across these areas - someone who can evaluate the acoustics of a room, rewrite an elicitation prompt, direct a pilot, and then turn what worked into a system that can be repeated hundreds of times.

Location

This role is based in Los Angeles and requires substantial in-person work.

Many of our collections involve hands-on studio prototyping, participants, physical environments, and production teams. Being present for pilots and collections is a core part of the role.