Ai Research Engineer (Agentic Post-Training)
Pubblicato il 15-10-2026 - Jobgether in Lazio
This position is posted by Jobgether on behalf of a partner company.
We are currently looking for an AI Research Engineer (Agentic Post-training) in Italy.
Si assicuri di inviare la sua candidatura con tutte le informazioni richieste, come indicato nella panoramica del lavoro riportata di seguito.
This role sits at the frontier of large language model development, focusing on advancing post-training techniques for agentic AI systems.
You will contribute to shaping models that go beyond text generation to actively reason, plan, and execute tasks through tool use and function calling.
The work spans research and engineering, with direct impact on production?grade AI systems deployed across real?world applications.
You will design and improve training pipelines that enable models to operate reliably in multi?step, multi?tool environments.
The environment is highly research?driven, collaborative, and fast?paced, bringing together experts in AI systems, multimodal learning, and distributed training.
Your contributions will directly influence the next generation of intelligent, autonomous AI agents capable of operating on both cloud and edge devices.
Accountabilities
Conduct end?to?end research and engineering work to advance post?training methods for agentic AI systems, focusing on tool use, reasoning, and autonomous behavior in real?world tasks.
Improve core model capabilities including factuality, instruction following, multi?step reasoning, tool/function calling, and multi?agent coordination.
Design, build, and optimize large?scale post?training pipelines, including data curation workflows, training infrastructure, and evaluation frameworks.
Develop robust benchmarking and diagnostic systems to assess model performance, reliability, and readiness for deployment.
Integrate real?world feedback signals from production usage into training loops to continuously enhance model behavior.
Collaborate closely with research, engineering, and product teams to ensure scalable,
production?ready integration of agentic capabilities.
Identify bottlenecks in current systems and propose novel solutions to improve efficiency, reliability, and performance of tool?augmented models.
Requirements
Degree in Computer Science, Machine Learning, or a related field; advanced degree (MS/PhD) strongly preferred.
Strong background in large language models, with proven experience in post?training techniques such as fine?tuning, reinforcement learning, or instruction tuning.
Hands?on experience with distributed training systems and large?scale model development (e.g., multi?GPU or multi?node environments).
Demonstrated expertise in improving model reasoning, tool use, function calling, or agentic workflows to achieve state?of?the?art performance.
Experience working with multimodal data (text, image, audio) and building or optimizing data pipelines for AI training.
Strong track record of research contributions, ideally including publications at top?tier AI conferences (e.g., NeurIPS, ICML, ICLR, ACL, CVPR, ECCV).
Open?source contributions related to AI agents, tool use, or LLM systems (e.g., GitHub, Hugging Face) is highly valued.
Strong analytical thinking, problem?solving skills, and ability to work in fast?paced, research?intensive environments.
Excellent communication skills and ability to collaborate effectively across technical and cross?functional teams.
Benefits
Remote?first and globally distributed working environment.
Opportunity to work on cutting?edge AI systems shaping the future of agentic intelligence.
Exposure to large?scale, real?world AI deployments and advanced research problems.
Collaborative environment with top?tier researchers and engineers across AI, systems, and product domains.
High?impact role with strong ownership over research and engineering initiatives.
Continuous learning and professional growth in a fast?evolving deep?tech ecosystem.
xlwpduy
Competitive compensation and performance?based growth opportunities (where applicable).
#J-*****-Ljbffr