Research Engineer – Applied Generative AI; LLMs & Multimodal Systems

Pocket FM is on a mission to deliver personalized and immersive audio experiences to listeners worldwide. We are revolutionizing the audio entertainment industry through long‑form storytelling, supported by our cutting‑edge platform that serves millions of listeners and generates billions of minutes of engagement monthly. We leverage Generative AI in producing content and streamlining operations, developing innovative solutions for cutting‑edge challenges in the AI landscape across all modalities—text, audio, and images.

With strong backing and rapid user base growth, Pocket FM is an exciting and dynamic place to join.

As a Research Engineer on our team, you won't just be experimenting—you'll be architecting the core of our AI‑first content stack. Your work will directly impact the stories our users hear every day.

• Go beyond notebooks. Design, build, and optimize robust pipelines for fine‑tuning foundation models (e.g., LLaMA, Mistral, Qwen) on custom datasets. You'll own the process from data curation and training to evaluation for tasks like narrative generation and dialogue synthesis.
• Engineer and product ionize models that seamlessly blend modalities. Your primary focus will be on creating state‑of‑the‑art systems for text, speech, and audio generation to power dynamic and immersive storytelling.
• Design and implement a scalable orchestration system (e.g., using Lang Graph, Ray, or custom frameworks) to manage complex, multi‑agent AI workflows. This includes planning agents, tool‑using models, and evaluation layers that work in concert.
• Bridge the gap between models and our production environment. You will integrate generative AI workflows with our cloud infrastructure, ensuring our systems for training, inference, and deployment are efficient, reliable, and scalable.
• Be the critical link between theoretical research and tangible product features. You'll read the latest papers, identify promising techniques, and write the high‑quality, efficient code needed to make them work at scale.

• You have a track record (ideally 4+ years) of building and shipping machine learning models into production environments. Your resume shows projects, not just publications.
• You have deep, hands‑on experience fine‑tuning large language models using techniques like LoRA, QLoRA, DPO, or RLHF. You can speak to the practical challenges and trade‑offs.
• You have practical experience building models that integrate multiple modalities (e.g., text‑to‑speech, audio understanding, vision‑language).
• You understand that a model is just one piece of the puzzle. You have experience building or maintaining the full AI pipeline, from data ingestion to model serving and monitoring.
• You have a bias for action and an obsession with shipping robust, efficient code. You know when to use an off‑the‑shelf solution and when to build from scratch.

• Proficiency in Python, PyTorch, and the Hugging Face ecosystem (Transformers, Accelerate, PEFT).
• Demonstrable experience with frameworks like FSDP, Deep Speed, or Megatron‑LM.
• Familiarity with AI workflow orchestrators (e.g., Lang Graph, Prefect, Ray) and experience connecting models to cloud infrastructure (AWS, GCP, or Azure).
• Experience with Retrieval‑Augmented Generation (RAG) systems, building AI agents, and designing novel evaluation frameworks for generative models.

• Work on unique, real‑world generative AI challenges that are defining the future of entertainment for millions of users.
• See your work go live and shape a product used by a massive global audience.
• We hire smart people and trust them. You'll have the freedom to drive your initiatives and choose the best tools for the job.

Join us in this exciting opportunity to contribute to ground breaking research in Generative AI technologies that will impact millions of users globally.