Job Title

Key Responsibilities Design and build LLM guardrails for prompt injection protection, toxicity/bias detection, and hallucination/jailbreak identification. Build and maintain evaluation frameworks to monitor LLM safety, fairness, and compliance. Develop automated pipelines to process, tag, and evaluate LLM outputs using Python and SQL. Leverage vector databases and embeddings to detect unsafe content or model drift. Create internal dashboards and visualizations (Streamlit, Dash, or lightweight React/JS) for POCs and internal tools. Collaborate with ML engineers and product teams to integrate LLM safety components into production APIs or applications. Stay current with AI safety research, emerging tools (Ragas, LangChain, Guardrails.ai), and regulatory standards (EU AI Act, NIST AI RMF). Required Qualifications 3 - 4.5 years of experience in data science, applied ML, or LLM-based applications. Strong programming skills in Python and experience writing SQL for data exploration or feature engineering. Solid understanding of NLP, deep learning (CNN/RNN/Transformers), and LLM architectures. Hands-on experience with Hugging Face, LangChain, LLM APIs (OpenAI, Anthropic), and vector stores (FAISS, Pinecone, Chroma). Familiarity with front-end basics (Streamlit, Dash, or simple HTML/CSS/JS) is aplus , not mandatory. Experience with model evaluation, red‑teaming, or safety interventions in NLP/LLM systems. (Preferred) Familiarity with deploying ML pipelines in production (Docker, FastAPI) and ability to thrive in a fast‑paced startup environment.