Master Internship - Tool‐augmented LLM agents for abdominal CT question answering: benchmarking, tool attribution and reasoning faithfulness

Inria Gif-sur-Yvette

Gif-sur-Yvette

On-site

EUR 8,900 - 16,000

Full time

3 days ago
Be an early applicant

Get a reply from this employer

Your resume and cover letter, tailored to exactly what they’re hiring for.

Get past ATS filters

Job summary

Inria Gif-sur-Yvette invites an internship to systematically evaluate LLM-based agents that combine reasoning with specialized medical imaging tools for abdominal CT question answering. The role focuses on accuracy, computational efficiency, tool contributions, and faithfulness of reasoning, in collaboration with APHP.

You will work with PyTorch, distant GPU servers, and multiple foundation models. The internship involves bibliographic study, programming, experiments, and scientific reporting,

Qualifications

  • Master 2 in ML or final year of engineering school.
  • Strong programming in Python and PyTorch with rigorous experimentation.
  • Experience with LLMs, Hugging Face, vLLM, and tool calling is a plus.

Responsibilities

  • Conduct bibliographic study relevant to LLM-based agents and medical imaging tools.
  • Implement experiments in PyTorch and manage GPU-based workloads remotely.
  • Perform ablation studies and benchmark agent-tool combinations.
  • Prepare scientific presentations and reporting of results.
  • Assess accuracy, efficiency, tool contributions, and reasoning faithfulness.

Skills

Python
PyTorch
LLMs
Hugging Face
vLLM
tool calling
agent frameworks
evaluation methodology

Education

Master 2 in machine learning

Tools

MONAI
DICOM
NIfTI

Job description

Context

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 models that 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 chest radiography 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 make up quantitative values.
The problem
: each system comes with its own harness (agent loop, tools, prompts), data and metrics, so it is impossible to tell 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.

Assignment

Research questions:

  1. 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. 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. 3. Is the agent’s reasoning faithful? When the agent justifies its answer, does it describe what actually determined it?
  4. Possible extension, depending on results and interest: the verifiable answers of DeepTumorVQA make it a natural testbed for reinforcement learning of agents [21] (e.g. GRPO on a small LLM).
Main activities
  • Bibliographic study
  • Programming in PyTorch
  • Use of distant GPU servers
  • Scientific presentations and reporting
  • Ablation studies and benchmarking
Skills
  • Master 2 in machine learning or final year of engineering school.
  • Good command of Python and PyTorch; taste for rigorous experimentation.
  • A plus: experience with LLMs (Hugging Face, vLLM, tool calling), agent frameworks or evaluation methodology.
  • No prior medical knowledge is required. A first experience with 3D imaging (NIfTI, DICOM, MONAI) is appreciated but can be acquired during the internship.
  • Fluent French or English; comfortable reading scientific papers in English.
Get your free, confidential resume review.

or drag and drop your file here.

Similar jobs

Similar jobs worth comparing

Master Internship - Tool‐augmented LLM agents for abdominal CT question answering: benchmarking, tool attribution and reasoning faithfulness
Master Internship - Tool‐augmented LLM agents for abdominal CT question answering: benchmarking, tool attribution and reasoning faithfulness

Inria • Gif-sur-Yvette

Hybrid
EUR 12,000 - 18,000
Transport reimbursement
Long leave allowance
Teleworking option
+2
Master Internship - Tool‐augmented LLM agents for abdominal CT question answering: benchmarking, tool attribution and reasoning faithfulness
Master Internship - Tool‐augmented LLM agents for abdominal CT question answering: benchmarking, tool attribution and reasoning faithfulness

Inria, the French national research institute for the digital sciences • France

Hybrid
EUR 13,000 - 17,000
Partial reimbursement of public travel
7 weeks of annual leave + RTT days
Teleworking possible
+2
Intern: Tool-Enhanced LLM Agents for Abdominal CT QA
Intern: Tool-Enhanced LLM Agents for Abdominal CT QA

Inria Gif-sur-Yvette • Gif-sur-Yvette

On-site
EUR 8,900 - 16,000
Master Internship: Tool-Augmented LLMs for Abdominal CT Q&A
Master Internship: Tool-Augmented LLMs for Abdominal CT Q&A

Inria • Gif-sur-Yvette

Hybrid
EUR 12,000 - 18,000
Transport reimbursement
Long leave allowance
Teleworking option
+2
Master Internship: Tool-augmented LLMs for Abdominal CT QA
Master Internship: Tool-augmented LLMs for Abdominal CT QA

Inria, the French national research institute for the digital sciences • France

Hybrid
EUR 13,000 - 17,000
Partial reimbursement of public travel
7 weeks of annual leave + RTT days
Teleworking possible
+2
Ml Internship - Learning from radiology reports to detect cancer on CT
Ml Internship - Learning from radiology reports to detect cancer on CT

StartX • Paris

Hybrid
EUR 12,000 - 21,000
Competitive salary
Premium computing hardware
Healthcare access
Research Intern
Research Intern

DeepHealth • Paris

Hybrid
EUR 17,000 - 28,000
Internship Proposal (M2 / Engineering Final Year) — PhD track
Internship Proposal (M2 / Engineering Final Year) — PhD track

Therapanacea • Paris

On-site
EUR 13,000 - 20,000
Machine Learning Internship - Multimodal Radiological Foundation Models
Machine Learning Internship - Multimodal Radiological Foundation Models

Raidium • Paris

Hybrid
EUR 20,311 - 24,328
Competitive salary
Access to premium computing hardware
Stage IA/ML Engineer 100% R&D - agentic AI
Stage IA/ML Engineer 100% R&D - agentic AI

Ekimetrics • Paris

Hybrid
EUR 18,000 - 22,000
Hands-on research experience
Collaborative learning
Skill development in LLM agents and AI
+2