TheLoops is looking for a Data Scientist to help shape the future of agentic AI systems by improving how AI agents reason, plan, and act in dynamic environments, and how they learn from real-world data, adapt to user intent, and interact autonomously or collaboratively.
Requirements
- Strong Python and SQL skills; experience with libraries like scikit-learn, PyTorch, LangChain or any similar agentic framework
- Familiarity with LLMs, retrieval-augmented generation (RAG), Reinforcement Learning, Fine-Tuning
- Experience working with agent frameworks (AutoGen, OpenAgents, LangGraph, etc.)
- Background in decision-making models, memory systems, or multi-agent coordination
- Exposure to vector databases, embeddings, and custom RAG pipelines
- Experience building evaluation frameworks or simulators for agent performance
- Experience with LLM Post Training SFT, DPO, RLHF, GRPO
Responsibilities
- Design and run applied research initiatives that inform the behavior and learning loops of AI agents
- Build and evaluate models for planning, memory, retrieval, reasoning, or tool use in agentic systems
- Develop internal tools and pipelines to test agent behavior across different environments
- Analyze structured and unstructured data from user-agent interactions, logs, and experiments
- Rapidly prototype and test hypotheses to improve agent performance and reliability
- Collaborate with engineering, product, and design to translate insights into deployable features
- Communicate findings clearly and concisely to both technical and non-technical audiences
Other
- Masters or PhD in Computer Science, Applied Math, Statistics, or related quantitative field
- 4+ years of experience in data science, applied ML, or AI research
- Comfort with ambiguity, fast iteration cycles, and self-directed research
- Excellent communication and storytelling skills you can explain complex models to others clearly
- Curious You explore the edges of what's possible with AI agents
- Action-biased You ship prototypes quickly and iterate based on evidence
- Scientific You bring structure to messy problems and back claims with data
- Independent You operate autonomously and drive projects end-to-end
- Collaborative You thrive in cross-functional teams and open discussion
