Data Scientist

Data Scientist

Join to apply for the Data Scientist role at Glia

Glia is the leading AI customer service solution for banks and credit unions. Our platform unifies AI and human agents across every voice and digital conversation through our proprietary Channel Less® Architecture. With AI for All™, organizations overcome the tradeoff between efficiency and experience by using AI to automate conversations and elevate service operations. Valued at over $1 billion and named a Deloitte Technology Fast 500™ company for five consecutive years, Glia powers over 700 financial institutions and maintains an industry-leading 72 NPS.

We're also certified as a Great Place to Work, with 98% employee satisfaction.

We are looking for a Data Scientist who has a passion for cutting edge Conversational AI. You won’t just be "using" models; you’ll be the architect of how our LLM-based virtual assistants learn, adapt, and perform. You will bridge the gap between cutting-edge research and production-grade stability, ensuring our GVAs (Glia Virtual Assistants) provide fast, human-like, and—most importantly—hallucination-free interactions for our clients.

• Architect & Maintain LLM Workflows:
  Design, implement, and optimize end-to-end pipelines for LLM-based chatbots, focusing on RAG (Retrieval-Augmented Generation) and agentic frameworks.
• Holistic Model Evaluation:
  Go beyond simple metrics. You will develop robust evaluation frameworks that measure:
  • Generative Output:
    Using LLM-as-a-judge, guardrails, and semantic similarity to ensure high-quality, on-brand responses.
  • Classifier Performance:
    Optimizing intent classification, sentiment analysis, and NER to ensure routing and logic remain flawless.
• Innovate with GVA Learning 360:
  Work on our proprietary technology that "clones" the wisdom of top human agents to automate complex banking scenarios.
• Applied Research:
  Act as a subject matter expert, staying ahead of the curve on breakthroughs in transformer architectures, reinforcement learning, prompt engineering and evaluation techniques.
• Identify Opportunities:
  Find areas where improvements can be made, such as additional capabilities (utilizing topic modelling), chaining models and agents to solve new workflows and increasing overall accuracy or reliability.
• Guardrail Engineering:
  Implement and refine safety layers to prevent hallucinations and ensure compliance within highly regulated industries (Finance).

• MSc. or PhD in computer science, mathematics, statistics, big data, or other data science disciplines
• 3-5 years experience working with large datasets and LLMs
• Product Focus with Feasibility Research & Strategic Analysis:
  Before diving into code, you perform rigorous feasibility studies. You can analyze the cost-benefit ratio of different approaches (e.g., "Is a specialized SLM more cost-effective than a general-purpose LLM for this specific task?") and assess the technical limitations of proposed products.
• Conversational AI Expertise:
  Proven experience building and deploying production-level conversational agents and NLP systems.
• LLM Proficiency:
  Deep knowledge of the LLM lifecycle—from prompt tuning and fine-tuning to managing context windows and latency optimization. Experience with Reinforcement Learning is a bonus.
• Evaluation Rigor:
  Experience with modern evaluation tools (e.g.,custom benchmarking suites or tools like lang Smith) and traditional ML metrics (F1-score, Precision-Recall).
• A Researcher Mindset:
  You are driven by a problem-solving mindset, testing hypothesis, questioning existing solutions and driving to improve great solutions and products.
• Technical Stack:
  Expert-level Python, experience with frameworks like PyTorch/Tensor Flow, and familiarity with vector databases (e.g. S3 Vectors or Pinecone) and LLM tools (e.g. Bedrock / Sagemaker).
• Data Preparation & Engineering:
  You possess a meticulous approach to data. You understand that the quality of a model is defined by the data it consumes. You are expert in
  • Cleaning and preprocessing unstructured conversational data (transcripts, chat logs).
  • Anonymizing Sensitive Personal Information (PII) to meet banking-grade security…