A complete application in a minute
Tailored resume and cover letter, ready to send.
Inria Saclay invites applications for a Master Internship focused on tool-augmented LLM agents for abdominal CT question answering. You will benchmark reasoning with medical imaging tools, analyze tool contributions and ensure faithfulness of reasoning, in collaboration with APHP.
The project reviews state-of-the-art CT models and asks you to systematically evaluate agents across accuracy and efficiency, with an emphasis on rigorous experimentation and reproducible results.
Created in 2008, the Inria Saclay Center is located at the heart of the Paris-Saclay scientific and technological excellence cluster, which alone accounts for 15% of French research. Serving the development of the Université Paris-Saclay and the Institut Polytechnique de Paris, the Inria Saclay center employs 80 people in research support services and 500 scientists of 54 nationalities.
Benefiting from continuous growth, the center now has a total of 42 project-teams and two in the process of being created, including 21 jointly with the Institut Polytechnique de Paris, 16 with the Université Paris-Saclay, as well as 7 Inria EPs, including one in collaboration with Onera and one with the Pôle Universitaire Centre Val de Loire. These research teams are spread over more than ten sites.
In the context of the ANR FM2AI, and in collaboration with APHP, the aim of this internship is to systematically evaluate LLM-based agents that combine reasoning with specialized medical imaging tools for abdominal CT question answering, focusing on their accuracy, computational efficiency, the contribution of individual tools, and the faithfulness of their reasoning.
Scientific context: Foundation models for CT have multiplied: vision‐language encoders (CT‑CLIP [10], Merlin [8], Jolia [15]), generative modelsthat answer questions about a volume (CT‑CHAT [10], MedGemma 1.5 [11]), large pretrained encoders (Pillar‑0 [12], Curia[13], FlexiCT [14]) and segmentation tools (TotalSegmentator [9]). On their own, they still fail on tasks that require localizing,measuring and then reasoning.
Hence the idea, following ReAct [16], of delegating the reasoning to an LLM that calls these models as tools. Born in chestradiography with MedRAX [7], this paradigm now reaches 3D CT: CT‑Agent [2], 3DMedAgent [5], CT‑Flow [6], INFORM‑CT [3].
The most striking result comes from RadAgent [1]: without any training, an agent orchestrating CT‑CHAT, CT‑CLIP, TotalSeg‑mentator and a general‑purpose 2D VLM (Gemma 3 [19]) outperforms CT‑CHAT on report generation. On DeepTumorVQA[4], a fine‑tuned 2D VLM even beats the best 3D model (66.3% vs. 59.8%), but without measurement tools, all models makeup quantitative values.
The problem: each system comes with its own harness (agent loop, tools, prompts), data and metrics, so it is impossible totell what actually makes the difference. Abdominal CT, less studied than the chest, is an ideal testbed: DeepTumorVQA pro‑vides questions with verifiable answers on 9,262 public volumes, organized by level (recognition, measurement, reasoning),together with a standardized tool environment.
Research questions:
1. What makes a good agent, and at what cost? Compare harnesses and open‑source LLMs run locally (Qwen3 [17], gptoss
[18], Gemma 4 [22]…) on the same task, and map the accuracy–cost frontier (tokens, latency, GPU memory).
2. Which tools really matter? Systematic ablations and attribution of performance to each tool (segmentation, measurements,
2D or 3D VQA such as MedGemma 1.5 [11], foundation models).
3. Is the agent’s reasoning faithful? When the agent justifies its answer, does it describe what actually determined it?
Possible extension, depending on results and interest: the verifiable answers of DeepTumorVQA make it a natural testbedfor reinforcement learning of agents [21] (e.g. GRPO on a small LLM).
Bibliographic study
Programming in PyTorch
Use of distant GPU servers
Scientific presentations and reporting
Ablation studies and benchmarking
This internship is strictly open to students in the final year of an engineering degree or enrolled in a second-year Master's programme (M2), preferably specializing in machine learning, artificial intelligence, or a closely related field. Applicants should have strong programming skills in Python and PyTorch, a solid background in machine learning, and an interest in rigorous scientific experimentation. Familiarity with large language models (LLMs), agentic AI, or evaluation methodologies would be an advantage. Prior experience in medical imaging is appreciated but not required. Applications from candidates outside the specified academic levels will not be considered.
Recruitment is subject to the candidate obtaining the required authorization to access the laboratory, which is classified as a Zone à Régime Restrictif (ZRR).