Senior Platform Engineer, Machine Learning
Fieldguide- Full Time
- Senior (5 to 8 years), Mid-level (3 to 4 years)
Candidates should have professional experience with model training, Huggingface Transformers, Pytorch, vLLM, TensorRT, and infrastructure as code tools like Terraform. Proficiency in scripting languages such as Python or Bash, and familiarity with cloud platforms like Google Cloud, AWS, or Azure is required. Experience with Git and GitHub workflows, as well as tracing and monitoring tools, is essential. Candidates should also be familiar with high-performance, large-scale ML systems and possess strong troubleshooting skills. A proactive approach to identifying problems and improving system performance is necessary, along with a commitment to building reliable and secure systems.
The Platform engineer, MLOps will work closely with AI/ML engineers and researchers to design and deploy a CI/CD pipeline for safe and reproducible experiments. They will set up and manage monitoring, logging, and alerting systems for extensive training runs and client-facing APIs. The engineer will ensure training environments are consistently available across multiple clusters and develop containerization and orchestration systems using Docker and Kubernetes. They will operate large Kubernetes clusters with GPU workloads, improve the reliability and quality of software solutions, and provide operational support for large-scale distributed software applications.
AI platform for business process optimization
Writer.com provides a platform that uses artificial intelligence to enhance business processes for various clients, including major companies. The platform integrates Large Language Models (LLMs), Natural Language Processing (NLP), and Machine Learning (ML) to tailor AI solutions to a client's specific brand and knowledge. LLMs generate human-like text, NLP helps computers understand human language, and ML enables learning from data. Writer's platform is built on secure, enterprise-grade LLMs called Palmyra, which are designed to be open and transparent, ensuring that client data is never used for training. This allows clients to self-host the models, giving them control and visibility over their data. The platform assists businesses in speeding up processes and generating customized outputs, such as summaries and insights, while adhering to legal and brand guidelines. Writer.com operates on a subscription-based model, where clients pay a recurring fee for access to its features.