Factspan - LLM Engineer

Job Overview

As an LLM (Large Language Model) Engineer, you will be responsible for designing, optimizing, and standardizing the architecture, codebase, and deployment pipelines of LLM-based systems. Your primary mission will focus on modernizing legacy machine learning codebases (including 40+ models) for a major retail clientenabling consistency, modularity, observability, and readiness for GenAI-driven innovation. Youll work at the intersection of ML, software engineering, and MLOps to enable seamless experimentation, robust infrastructure, and production-grade performance for language-driven systems. This role requires deep expertise in NLP, transformer-based models, and the evolving ecosystem of LLM operations (LLMOps), along with a hands-on approach to debugging, refactoring, and building unified frameworks for scalable GenAI :

• Lead the standardization and modernization of legacy ML codebases by aligning to current LLM architecture best practices.
• Re-architect code for 40+ legacy ML models, ensuring modularity, documentation, and consistent design patterns.
• Design and maintain pipelines for fine-tuning, evaluation, and inference of LLMs using Hugging Face, OpenAI, or open-source stacks (e.g., LLaMA, Mistral, Falcon).
• Build frameworks to operationalize prompt engineering, retrieval-augmented generation (RAG), and few-shot/in-context learning methods.
• Collaborate with Data Scientists, MLOps Engineers, and Platform teams to implement scalable CI/CD pipelines, feature stores, model registries, and unified experiment tracking.
• Benchmark model performance, latency, and cost across multiple deployment environments (on-premise, GCP, Azure).
• Develop governance, access control, and audit logging mechanisms for LLM outputs to ensure data safety and compliance.
• Mentor engineering teams in code best practices, versioning, and LLM lifecycle Skills :
• Deep understanding of transformer architectures, tokenization, attention mechanisms, and training/inference optimization
• Proven track record in standardizing ML systems using OOP design, reusable components, and scalable service APIs
• Hands-on experience with MLflow, LangChain, Ray, Prefect/Airflow, Docker, K8s, Weights & Biases, and model-serving platforms.
• Strong grasp of prompt tuning, evaluation metrics, context window management, and hybrid search strategies using vector databases like FAISS, pgvector, or Milvus
• Proficient in Python (must), with working knowledge of shell scripting, YAML, and JSON schema standardization
• Experience managing compute, memory, and storage requirements of LLMs across & Experience :
• 5+ years in ML/AI engineering with at least 2 years working on LLMs or NLP-heavy systems.
• Able to reverse-engineer undocumented code and reimagine it with strong documentation and testing in mind.
• Clear communicator who collaborates well with business, data science, and DevOps teams.
• Familiar with agile processes, JIRA, GitOps, and confluence-based knowledge sharing.
• Curious and future-facingalways exploring new techniques and pushing the envelope on GenAI innovation.
• Passionate about data ethics, responsible AI, and building inclusive systems that scale

(ref:hirist.tech)