Pocket FM is building the world's most advanced AI copilot for fiction writers, helping them go from a blank page to a fully produced audio series across web and mobile. We build AI that understands narrative structure, character arcs, pacing, emotional continuity, and story evolution across hundreds of episodes, not a generic chatbot. With 100M+ listeners globally, we sit at the intersection of generative AI, storytelling, and entertainment infrastructure.
What you'll build
- Planner: story planning systems for characters, world building, plot arcs, and pacing; involves narrative reasoning, long context planning, agentic workflows, and memory/knowledge graphs
- Prose: scene generation, dialogue, prose refinement, and tone adaptation using long context, controllable, and multilingual generation models
- Feedback: AI editing systems that catch pacing, continuity, and structural issues like a developmental editor, using critique and evaluation models with human in the loop workflows
- Audio Native Infrastructure: voice over formatting, audio mastering, trailer generation, and narration aware adaptation
Who we're looking for
- 3+ years building ML/LLM systems in production
- Strong applied research experience in Generative AI, with hands on LLM fine tuning and post training
- Experience building AI copilots, agents, or workflow assistants
- Experience translating research into production systems, especially long context, memory, planning, or evaluation
- Exposure to creative or content generation systems is a strong plus
- Genuine curiosity about content, storytelling, or creative tooling
Multilingual & Hindi LLM experience
- Hindi or multilingual/non English LLMs, fine tuning multilingual foundation models, instruction tuning, domain adaptation, cultural and linguistic alignment, and low resource language optimization
- Experience with Indian language AI systems is highly valued
Technical Stack
- Required: Python, PyTorch, Hugging Face, distributed training (DeepSpeed, FSDP, Megatron LM), production AI systems experience
- Preferred: LangGraph/Ray, RAG and memory systems, AI evaluation pipelines, multilingual NLP/Hindi LLMs, speech/audio AI
