Senior Generative AI Engineer — Founding Team, Nexyll Studios (Pexyl AI)

About the role

Pexyl AI is building Nexyll Studios — a proprietary AI-native video production platform designed to produce world-class micro-series, edutainment content, music videos, and product ads at a fraction of traditional production cost and time. We're not building a wrapper around a single model — we're building the orchestration layer that turns a script into a finished, broadcast-quality episode by composing multiple best-in-class generative systems into one pipeline.

We're hiring a founding Senior Generative AI Engineer to design and own that pipeline end to end: from prompt/scene decomposition through video, image, voice and music generation, to final assembly — at production scale (50+ episodes and growing).

What you'll own

- Design and build the core content-generation pipeline that orchestrates multiple generative APIs (video, image, voice, music) into a single automated production workflow — from script/shot-list generation through final render.
- Architect this orchestration using MCP (Model Context Protocol) and agentic tool-calling patterns, so the pipeline can compose and swap models (video generation, image generation, TTS, LLMs) as tools rather than hardcoded integrations.
- Build and tune the character/scene-consistency layer — reference-image conditioning, LoRA fine-tuning, and prompt engineering — so recurring characters stay visually consistent across dozens of episodes.
- Own model selection and cost/quality tradeoffs across the generative stack (e.g., video generation models like Seedance/Kling/Sora-class systems, image models like Nano Banana/Nano Banana Pro, avatar systems like HeyGen, voice/music systems like ElevenLabs) — benchmarking quality, latency, and per-unit cost,



and building fallback/routing logic between them.
- Build the GPU/inference infrastructure where needed — evaluating when to use managed APIs versus self-hosted open-weight models (e.g., for LoRA training or high-volume inference), and optimizing utilization on rented GPU infrastructure (RunPod, Lambda, or equivalent).
- Instrument the pipeline for quality control at scale: automated evaluation of generated shots, retake/regeneration logic, and human-in-the-loop review checkpoints.
- Work directly with founders and the creative team to translate storytelling and production needs into pipeline capabilities — this is a build-the-plane-while-flying-it role.
- Own infra cost discipline: you'll be accountable for the unit economics of every generated minute of content, not just whether the pipeline works.

What you bring

- 2+ years working hands-on with generative AI systems (diffusion models, video generation, LLM orchestration, or multimodal pipelines) in production, not just research/notebooks.
- Strong Python engineering fundamentals; experience building reliable, async, API-orchestration systems (retries, queuing, rate-limit handling, cost tracking across multiple third-party AI providers).
- Practical experience with MCP or comparable agentic tool-orchestration frameworks — you understand how to compose LLMs and generative tools into multi-step, self-correcting workflows, not just call a single API.




- Hands-on experience with at least one contemporary video/image generation stack (Seedance, Kling, Runway, Sora, Stable Diffusion/ComfyUI, or similar) and an understanding of their prompt-conditioning and reference-image mechanics well enough to push past demo-level output.
- Experience with voice/TTS pipelines (ElevenLabs or similar) and audio-video sync.
- Working knowledge of GPU infrastructure economics — when self-hosting beats managed APIs, how to size and optimize inference on rented GPU capacity (H100/A100 class), and basic LoRA/fine-tuning workflows.
- Comfort operating with a cost-per-unit mindset — you think in cost-per-second-of-video and cost-per-episode, not just "does it work."
- Prior 0-to-1 or early-stage startup experience, ideally in media-tech, content-tech, or consumer AI products.

Nice to have

- Direct experience in the microdrama/vertical-video/short-form content space, or with edutainment or ad-generation platforms.
- Experience building character-consistency systems (LoRA, DreamBooth-style fine-tuning, IP-Adapter/reference-conditioning) for serialized content.
- Familiarity with video post-production automation (FFmpeg pipelines, automated editing/assembly, upscaling).
- Published work, open-source contributions, or demos in generative video/audio.

Why join You'll be the founding technical owner of the generation pipeline for a category-defining AI studio — not iterating on someone else's architecture, but building it from the ground up, with direct input into which models the entire company's content quality and unit economics depend on. Meaningful equity, high autonomy, and a front-row seat as the AI micro-drama/edutainment category scales globally.

📌 Lead Generative AI Engineer (Mumbai)
🏢 Pexyl
📍 Mumbai