🌐 Welcome to TANITAI

Tanit AI is a healthcare technology startup pioneering neuro-symbolic AI to transform fertility and reproductive medicine. In a world where 1 in 6 people are affected by infertility, our mission is to elevate fertility practice for doctors and empower patients in their parenthood journey

➡ Mission

At Tanit AI, our mission is to leverage advanced artificial intelligence to educate and support individuals in their fertility journey.

We strive to provide accessible, personalized information and guidance to help people make informed decisions about their reproductive health.

➡ Vision

Our vision is to be a global leader in AI-driven fertility education, fostering a future where technology and human understanding converge to address critical reproductive health challenges. We aim to empower individuals worldwide with the knowledge and tools needed to navigate their fertility journey with confidence.

Role Description

We are looking for a Knowledge Graph Reasoning & Deep Learning Intern to explore how knowledge graphs and graph neural networks can enhance the factuality, interpretability, and multi-hop reasoning capabilities of LLMs in reproductive medicine.

You will work at the intersection of Neo4j, biomedical ontologies, GNNs, and LLM-based RAG, designing graph-aware reasoning pipelines to support complex clinical questions in fertility care.

Key Responsibilities

🕸️ Graph schema & reasoning design

• 🧩 Design and refine the schema of a medical knowledge graph focused on fertility (concepts, relations, hierarchies).
• 🔍 Implement multi-hop reasoning over the graph using Cypher queries and/or graph algorithms (shortest paths, k-hop neighborhoods, influence propagation).

🧠 Deep learning on graphs & GraphRAG

• 🧬 Explore Deep Learning on Graphs (GNNs) for tasks such as risk stratification, node classification, or link prediction.
• 🛰️ Build a GraphRAG pipeline where LLMs query the graph and a vector store to answer medical questions.

✅ Factuality, fact-checking & explainability

• 🛡️ Experiment with fact-checking layers that verify LLM outputs against the graph and flag contradictions or missing facts.
• 📊 Evaluate factuality, coverage, and explainability of graph-augmented LLMs compared to baseline pure-LLM approaches.
• 🔎 Surface subgraphs and reasoning paths to provide transparent explanations for clinical answers.

Required Skills

• 🐍 Strong Python skills (PyTorch or TensorFlow, data processing, APIs).
• 🗃️ Familiarity with graph databases (Neo4j) and basic graph theory.