Responsibilities
- Develop and integrate RAG, AI agents, recommendation and search systems, and AI-powered automation workflows.
- Build and maintain data ingestion, embedding, retrieval, and evaluation pipelines.
- Collaborate with developers and QA to design, build, test, and deliver AI features.
- Work with DevOps to deploy, monitor, and optimize self-hosted LLMs.
- Create evaluation gates for LLM changes and continuously improve accuracy, latency, cost efficiency, security, and reliability.
Requirements
- At least 3+ years of relevant AI/ML engineering experience.
- Strong Python skills, including building data pipelines for training datasets.
- Understanding of prompt engineering, RAG, embeddings, vector databases, tool calling, AI agents, and LLM evaluation.
- Experience with APIs, SQL and NoSQL databases, Docker, Git, and production AI systems.
- Experience with LLM serving and APIs such as OpenAI, Anthropic, or Azure OpenAI.
- Experience with inference tools such as vLLM or Ollama, including quantization and memory, latency, and quality trade-offs on constrained hardware.
- Ability to design and run evaluations for nondeterministic systems and quantify improvements.
- Strong English verbal and written communication skills.
- Preferred qualifications include open-weight LLM fine-tuning, agent frameworks, LLM security, on-premises or offline compact-model deployment, evaluation tooling, and cybersecurity experience.
Benefits
- The position is onsite.
- OPSWAT is an equal opportunity employer.
Categories
About OPSWAT
OPSWAT protects critical infrastructure. Our goal is to eliminate malware and zero-day attacks. We believe that every file and every device pose a threat. Threats must be addressed at all locations at all times—at entry, at exit, and at rest. Our products focus on threat prevention and process creation for secure data transfer and safe device access. The result is productive systems that minimize risk of compromise. That’s why 98% of U.S. nuclear power facilities trust OPSWAT for cybersecurity and compliance. OPSWAT. Trust no file. Trust no device. www.opswat.com Visit us on Twitter, Facebook, Instagram, and YouTube. http://www.twitter.com/opswat http://www.facebook.com/opswat https://www.instagram.com/opswat https://www.youtube.com/user/opswat1