All open roles

CAREERS AT LYANAI

Algorithm Engineer — LLMs & Affective Computing

LLMAffective computingModel trainingEvaluation

About the role

We are exploring how large language models can support real human communication, understanding not only words but also context, emotions, relationships and intent.

We are looking for an algorithm engineer interested in large language models, affective computing and human-computer interaction.

You will help build LYANAI’s core AI capabilities and improve performance in real conversations through data, training, evaluation and inference optimisation.

The role goes beyond calling model APIs and involves more than purely academic research.

We aim to turn valuable research methods into useful product capabilities.

Responsibilities

  1. Improve large language models for dating and social communication scenarios.
  2. Research emotion understanding, communication intent, context modelling and expression strategies.
  3. Build training data, clean datasets, assess quality and develop preference datasets.
  4. Run SFT and preference optimisation experiments, and analyse their practical results.
  5. Build model evaluation systems to improve reply quality, naturalness, relevance and appropriateness.
  6. Develop AI application capabilities including agents, context management and inference optimisation.
  7. Follow research in LLMs, affective computing and related fields, and explore methods suited to our product.

Requirements

  1. Foundations in machine learning, deep learning or natural language processing.
  2. Familiarity with Python, PyTorch and common development tools.
  3. An understanding of Transformers, large language models and common training and inference methods.
  4. Knowledge of fine-tuning techniques such as SFT, LoRA and preference optimisation.
  5. Skills in data processing, experiment design, model evaluation and analysis.
  6. The ability to read technical documentation and research papers in English.
  7. An interest in affective computing, natural language understanding or human-computer interaction.
  8. Strong practical skills, with the ability to implement and validate algorithmic ideas.

Nice to have

  • Practical experience with LLM fine-tuning, preference learning or reinforcement learning.
  • Familiarity with Hugging Face, Transformers, PEFT and TRL.
  • Knowledge of vLLM, model quantisation or inference deployment.
  • Project experience in sentiment analysis, dialogue systems or user intent recognition.
  • An interdisciplinary background in psychology, cognitive science or computational linguistics.
  • Experience developing agents, LangGraph applications or model evaluation systems.
  • Relevant publications, open-source projects or independent research.

What we value

Research and engineering together

Understand the algorithms, design experiments, test hypotheses and apply the results to real products.

Practical impact

Metrics provide guidance. We care about whether a model improves the user’s communication experience.

A spirit of exploration

There is much to explore where affective computing meets large language models. We welcome people from different disciplines who take the initiative to investigate new methods.

Data and experiments

Every model improvement should have traceable experimental evidence, beyond subjective impressions.

What you will gain

  • A direct role in building the core model capabilities of a real AI product.
  • Experience across the full cycle of data, training, evaluation, deployment and feedback.
  • Opportunities to explore the intersection of LLMs, affective computing and dialogue systems.
  • Participation in discussions on training methods, evaluation systems and technical direction.
  • Space to explore and put ideas into practice in a small team.

Let’s turn ideas into products.

Passionate about AI, affective computing and human-computer interaction?
Join LYANAI to explore how technology can support human connection.

Send your CVyhuangkeji@163.com
Other open rolesFull-stack Engineer Other open rolesData Annotation Engineer