Vultron is applying generative AI and agentic systems to government contracting, aiming to improve model quality in real-world workflows, develop custom evaluation and feedback loops, and balance performance, scale, and control for open and closed models.
Requirements
- Experience fine-tuning and prompting large models particularly agentic or task-specific systems and improving performance through grounding, context window optimization, or RAG.
- Experience researching and building LLM-based systems, including both training and deployment.
- Experience with scalable LLM systems using LangChain, LlamaIndex, Hugging Face, or similar frameworks.
- Familiarity with GPU-based experimentation and cloud model deployment (e.g., vLLM, DeepSpeed, Ray Serve)
- Experience with RAG pipelines and evaluation frameworks like RAGAS, DSPy, or LLaMAEval.
- Significant contributions to the AI field via top-tier conferences or open-source projects.
- 5+ years of experience in ML research, applied LLMs, or model infrastructure.
Responsibilities
- Build and improve model-powered product features that help federal contractors reason faster and act with confidence.
- Enhance grounding, summarization, and generation systems across Vultrons modular proposal workflows.
- Develop and refine retrieval systems using RAG, vector search, and knowledge graphs.
- Design evaluation pipelines that go beyond BLEU or ROUGE measuring utility, accuracy, and synthesis quality in noisy, complex documents.
- Conduct experiments with fine-tuning, synthetic data generation, and in-context learning to improve performance across retrieval and summarization tasks.
- Improving synthesis, summarization, grounding, and structured reasoning in a government contracting and proposal context.
- Designing robust ways to evaluate relevance, completeness, hallucination risk, and user-centric performance.
Other
- Must be located in the SF Bay Area or be willing to relocate.
- Proven ability to work closely with engineers, product teams, and non-technical experts to ship fast and stay aligned.
- Background as an early engineer in a rapidly growing startup.
