Website:
logline.ai
Job details:
The Role
This is a research role with a short path to production. You won't be writing papers for their own sake — you'll be investigating problems that matter to how creative professionals use AI, and translating what you learn into systems and techniques that ship. The line between research and engineering at LoglineAI is deliberately blurry: if you discover something that works, you're expected to help make it real.
The right person here has strong research fundamentals, genuine curiosity about the intersection of language models and creative cognition, and the discipline to move from interesting finding to usable output without losing rigour along the way. Experience in a formal research institution is valued but not required — what matters is the quality of your thinking and the relevance of what you've built or published.
What You'll Do
- Investigate and advance LoglineAI's core research questions: how LLMs can support long-form narrative development, how creative quality can be evaluated systematically, how cultural and linguistic context shapes model behaviour in a Bengali and Indian entertainment setting, and how human-AI creative collaboration can be designed to produce better stories.
- Design and run experiments to evaluate model capabilities, failure modes, and performance boundaries across creative tasks — developing rigorous evaluation frameworks that go beyond perplexity and BLEU scores.
- Explore and adapt state-of-the-art techniques — fine-tuning, RLHF, preference learning, RAG, agentic architectures — and assess their applicability to LoglineAI's specific creative and production use cases.
- Build proof-of-concept implementations that demonstrate research findings in a form engineering teams can act on — not just write-ups, but working code that shows the idea functioning under realistic conditions.
- Monitor the research landscape continuously: read papers, track model releases, follow open-source developments, and surface what's relevant to the team before it becomes common knowledge.
- Work directly with writers, directors, and creative teams to understand the qualitative dimensions of creative quality that quantitative metrics miss — and find ways to operationalise those insights into model evaluation and training.
- Contribute to LoglineAI's internal knowledge base: document research findings, experimental results, and technique assessments in a form that non-researchers on the team can use and build on.
- Collaborate with the engineering team to transition research outcomes into production features — staying involved through integration, not just handing off a notebook.
Must-Have Qualifications
- 3–5 years of experience in AI or ML research — either in an academic, industry lab, or applied research setting — with work you can point to that shows rigorous experimental design and honest evaluation of results.
- Deep familiarity with large language models: pre-training, fine-tuning, RLHF, RAG, prompt engineering, and evaluation methodologies — you understand the mechanics well enough to design experiments that actually test what you think they're testing.
- Proficiency in Python and standard ML tooling (PyTorch or JAX, HuggingFace, experiment tracking tools) — you can implement, run, and iterate on experiments independently.
- Strong written communication: you can write a research brief, an experimental debrief, and a plain-language summary of a complex finding — all three clearly and at the appropriate level of detail for the audience.
- AI-native research practice: you use AI tools to accelerate literature review, code generation, and synthesis — and you apply the same critical rigour to AI-generated outputs that you would to any other source.
- Genuine intellectual curiosity about narrative, creativity, and storytelling — not just about AI as a technical field. The problems here live at the intersection of both.
Good-to-Have
- Published or publicly shared work — papers, blog posts, open-source projects, or talk recordings — that demonstrates your thinking and research quality. Venue prestige matters less than the quality of the work itself.
- Experience working with multilingual or low-resource language models
- Familiarity with creative writing, screenwriting, or narrative theory — either as a practitioner or as a serious reader — that gives you intuitions about what makes storytelling work beyond what a loss function can capture.
- Exposure to evaluation techniques for open-ended generative tasks: human preference studies, comparative evaluation frameworks, or model-based evaluation using LLM judges.
Click on Apply to know more.