Principal AI Engineer, Media Pipelines

About Liquid Studios

Liquid Studios turns existing video, audio, and text catalogs into platform-native short-form clips, multi-language dubs, and formats that AI search and assistants can find and cite. We run it as a managed service that combines multi-model AI orchestration, agentic pipelines, and enterprise-grade quality control. Our clients are creators and media brands that want their catalogs to reach more platforms, in more languages.

Position Overview

Liquid Studios is seeking a highly technical, hands-on, and AI-native Principal Engineer to drive our product roadmap and lead execution on our core content transformation pipeline.

We're hiring a Principal Engineer to own the architecture and technical direction of the pipeline that turns media catalogs into short-form clips, multi-language dubs, and AI-discoverable formats. This is a senior individual contributor role. You’ll partner closely with our Lead AI Engineer and cross-functional leads to set the technical direction and architectural standards for the pipeline.

The hard problems: orchestrating video, speech, and text models across entire catalogs; keeping inference cost and latency in check at that scale; and building evaluations that tell us whether a dub or a clip is good enough to ship

What You’ll Own

• Pipeline Architecture: Set the technical direction for how media moves through ingestion, transformation, quality control, and delivery. Make the calls on orchestration, data models, and service boundaries, and document them so others can build on them.

• Model and Vendor Strategy: Decide what to build, buy, or fine-tune across LLMs, speech recognition, text-to-speech, translation, and video models. Run fast, structured bake-offs when new models ship, and design the pipeline so providers can be swapped without a rewrite.

• Cost, Latency, and Throughput: Own the unit economics per minute of processed media. Manage GPU and API spend through batching, caching, and routing, and degrade gracefully when a provider is slow or down.

• Evaluation and Quality: Build evals for subjective output such as dub naturalness and sync, translation accuracy, clip selection, and brand compliance. Turn them into automated gates (verification, silence or failure detection) backed by targeted human review.

• Enterprise Reliability: Define SLOs, observability, incident response, and data handling for client catalogs that are often unreleased or rights-restricted.

• Rights, Consent, and Safety: Work with leadership on voice-cloning consent, likeness, content rights, and AI disclosure, and build those requirements into the pipeline from the start.

• Engineering Standards: Set the bar for design docs, code review, and testing. Mentor other engineers through pairing and review.

• Hands-On Delivery: Write a meaningful share of the critical code yourself, and take high-priority roadmap features from prototype to production.

What Success Looks Like

In your first 90 days

• You know the pipeline end to end and have shipped at least one high-priority roadmap feature to production.

• You've written an architecture review: where the pipeline breaks at 10x volume, and the plan to fix it.

• You've run a structured model bake-off and changed a production model choice based on the results.

In your first 12 months

• Cost-per-processed-minute is meaningfully lower, and you can show where the savings came from.

• Every output type has automated quality gates and an eval score we track release over release.

• Adding a new model, language, or output format takes days, not weeks.

• Enterprise clients get the reliability they'd expect from a mature vendor: clear SLOs, incident reporting, and no silent failures.

How You Work

• AI-Native & Experimental Mindset: You experiment constantly. You test new models as they come out, build your own tools and agents, and have prototypes you can show us. You also know when a prototype is ready for production and when it isn't

• Intrepidness & Audacity: You're comfortable without a playbook. Much of what we build hasn't been done before. You make sound bets on incomplete information, say what would change your mind, and change course when it does.

• Humility & Integrity: Exceptional collaboration skills without ego. Highly receptive to feedback, team-oriented, and deeply respectful of both technical and creative disciplines.

• Product-Minded Builder: Highly execution-focused; you build with clear commercial utility and user impact in mind rather than just experimenting in isolation.

• Self-Directed & Proactive: You own outcomes. You work well remotely, set priorities against company goals, and keep others informed without needing day-to-day oversight.

Qualifications & Experience

Required:

• Proven Engineering Experience: Senior engineering track record: Typically 10+ years building software, including time as the technical owner of a production system with real users, uptime commitments, and cost constraints.

• Generative & Multimodal AI Proficiency: Production generative AI: You've shipped LLM or multimodal systems to production, such as agentic workflows, multi-step pipelines, or RAG. You've handled the parts demos skip: evals, retries, failure modes, and cost.

• Modern stack fluency: Deep experience in Python and/or TypeScript and at least one major cloud platform. You choose tools on merit and can explain the tradeoffs.

• Distributed systems and data : Strong grounding in job orchestration, queues, storage, and databases for large, long-running, failure-prone workloads.

• Technical influence: You've set direction for a system other engineers built on, through design docs, reviews, and mentoring.

Nice to Have

• Media Processing Fundamentals:

  • Media processing: FFmpeg, codecs and containers, transcoding, loudness normalization, and caption or subtitle formats.

  • Speech and audio ML: speech recognition, text-to-speech, voice cloning, speaker diarization, or lip-sync.

  • Video understanding: scene detection, highlight and clip selection, or multimodal embeddings.

  • Self-hosting open-source models on GPUs, including capacity and cost planning.

  • Retrieval and vector search over large content catalogs.

  • Experience in media, publishing, podcasting, or creator platforms, or selling to enterprise media clients.