Knowledge Engineer

While Xiaomi MiMo Team as a whole is committed to building an intelligence that understands the world, the Knowledge Engineer Unit is dedicated to sharpening its cognitive core. Unsatisfied with merely imitating patterns in massive datasets, we question the real meaning of understanding.

The trajectory of human knowledge suggests some common structures of reasoning and paradigms of cognition: mathematical deduction, philosophical dialectics, and metaphorical representation...

We draw upon the literature of diverse knowledge to shape the model's way of thinking. The realisation of AGI is not merely about answering existing questions. It is about enabling them, like experts from various disciplines, to actively apply knowledge, discover knowledge, reorganize knowledge, and create knowledge in a broad, open world.

Like a Renaissance liberal mind, we draw on multiple disciplines to reach a deeper understanding. And it is the depth of thought, that defines the altitude of intelligence.

As a member of the Xiaomi MiMo Knowledge Engineer Unit, you will work on innovative research in reasoning enhancement for large models. We do not confine members with rigid "job descriptions"; instead, we collaborate around the following core questions:

• Building cross-disciplinary reasoning — How can we enable models to carry out rigorous logical inference in mathematical and scientific fields, while at the same time forming insightful analysis and judgments in the humanities and social sciences?
• Redefining scientific benchmarks — Do current evaluation benchmarks measure "intelligence" or "tricks"? We aim to develop new benchmarks for scientific reasoning that go beyond score-chasing and do assess a model's ability in analyzing and solving complex academic problems.
• Distilling high-quality knowledge — A model's cognitive depth is rooted in the quality of the knowledge it absorbs. We work on how to efficiently discover, distill, construct, and structure high-quality disciplinary knowledge, building a solid ladder for the model's cognitive advancement.

We are looking for companions who are not only experts in a single field, but also curious about and respectful of knowledge itself, and who enjoy exploring across boundaries:

• Disciplinary background — Systematic academic training (bachelor's, master's, or PhD) in a given field (literature, philosophy, psychology, mathematics, physics, chemistry, biology, medicine, engineering, etc.), together with a strong curiosity and appetite for knowledge in other disciplines.
• AI and engineering skills — Proficiency in Python, interest in large-model technologies, and a desire to turn disciplinary insight into tangible, iteratable model capabilities.
• Research capacity — A rigorous academic attitude, a remarkable capacity in logical reasoning, and data analysis. Whether you bring high-quality publications in relevant fields or hands-on experience in solving complex problems, we value both.

• Inter-disciplinary thinking and innovation — Our team offers a superb cross-disciplinary atmosphere where sparks of thought are part of our daily routine. You will work side by side with top engineers, Olympiad medalists, and preeminent peers from diverse academic backgrounds to drive innovative breakthroughs in frontier technologies.
• Impactful exploration — We believe in pragmatic building. With "Xiaomi speed", your research outcomes will not stop within theory; they will be rapidly implemented and iterated in our open-source models, witnessed, used, and co-created by developers worldwide, and will concretely promote exchange and progress across the industry.
• Self-growth and fulfillment — Here you will confront the most central questions in today's AI studies. This challenging process will push you beyond your limits, reshape your research perspective, and refine your comprehensive capabilities. The value created by your work will accumulate as the cognitive depth of the next generation of large models and be seen by the world.

• MY : I'm not doing something repetitive. I can feel that I'm touching the limits of what our models can do and exploring how they can get even better. Creating things gives me a real sense of achievement, and I'm lucky to work side by side with teammates who are warm and amazing. And it is a truly one-of-a-kind experience for me to witness how our newborn MiMo is gradually growing better and stronger.

• LTC : To me, being a knowledge engineer here doesn't feel like just a job; it's much more of a chance to keep learning and keep creating. Here, large models are both the products and the tools. Knowing the LLMs' limits gives me a sense of control. And I've learned how to use AI to reach a level of productivity hitherto undreamed of. At one point, the rapid progress of LLMs made me suspect the meaning of my academic training. But now I'm convinced that the humanities' way of thinking is the hardest thing to replace, and in the era of AI, endless possibilities are to be realised by the humanities.

• MC : As a hardcore sci-fi fan, I just can't say no to this work that brings the future into the present. This job has earned me plenty of strange looks from people in my field, and as models grow stronger, doubts about my own value as a filmmaker come knocking again and again. But I shall never retreat in front of uncertainty, especially when I'm offered the best team I've ever worked with. In fact, working across different fields just isn't a problem as long as everyone enjoys the inspiration. Fate is meant to swing between existential anxiety and the shiver of creation. Who could resist that latter's pull? I truly believe we are among the first people to step through this gate of light, bringing the future into the present. The forthcoming is coming.

• WY : In old wuxia stories, the "Knowledge Engineer" was the one who knew everything happening in the martial world. But no one could carry all that knowledge alone. So today's "Knowledge Engineer" has become something new — a shared, collaborative place where people pool what they know. Here, everyone brings what they're good at. Everyone gets what they need. We trade insights, bridge each other's gaps, and support the travelers who move through the world of large models. If you're someone who wants to understand the field, connect with the community, and stay ahead of the curve — this is your place. Here you can enjoy the joy of spotting things early, meeting great people from all corners, and shaping the future through data. Come join us. Let's build something big together and push toward AGI.

• SR : Before I arrived, I carried a quiet worry — the same one many people in the humanities have. Large models were improving so quickly that I doubted how much space was left for me. But after joining the Knowledge Engineer Group and working with the models every day, the worry softened, and in its place, a quiet kind of wonder slowly bloomed. I no longer saw the work as a kind of "technology takes over." Instead, it is "collaboration". When we guide the model's thinking, its responses push us to examine our own assumptions — especially what we really mean by "understanding." And somehow, this question deepened my sense of the humanities instead of weakening it. Here, human instinct meets machine logic. We guide the model. But we are also changed by it. The more models evolve, the more I believe in the value of human imagination — our ability to break patterns, capture ideas, and create from the ground up. Technology isn't here to replace that. It gives us more space to grow. That, to me, is the purpose of Knowledge Engineer, and the direction we're heading.