Abhay Puri
Applied Research Scientist
Agentic Harness & Defenses
I am a Staff Research Engineer/Scientist at ServiceNow AI Research, working on the safety and security of LLMs and AI agents red-teaming, backdoor attacks, adversarial fine-tuning, prompt injection, and secure adaptation for enterprise systems. I have also worked on multimodal learning, document understanding, and scalable vector graphics (SVG) generation. I completed my M.Sc. in Computer Science (specialization in ML) at Mila / Université de Montréal. My work has appeared at NeurIPS, ICLR, CVPR, COLM, TMLR, and CAIS. I’m always happy to chat about AI security, agent robustness, or research in general – feel free to reach out!
Intérêts
- Computer Vision
- Natural Language Processing
- Large Language Models
- Trustworthiness
- Diffusion Models
- Machine Learning
Publications
Malice in Agentland: Down the Rabbit Hole of Backdoors in the AI Supply Chain.
ACM Conference on AI and Agentic Systems,
2026.
No, of Course I Can! Deeper Fine-Tuning Attacks That Bypass Token-Level Safety Mechanisms.
International Conference on Learning Representations,
2026.
AlignVLM: Bridging Vision and Language Latent Spaces for Multimodal Document Understanding.
Neural Information Processing Systems (NeurIPS),
2025.
Rendering-Aware Reinforcement Learning for Vector Graphics Generation.
Neural Information Processing Systems (NeurIPS),
2025.
BigCharts-R1: Enhanced Chart Reasoning with Visual Reinforcement Finetuning.
Conference on Language Modeling (COLM),
2025.
DoomArena: A framework for Testing AI Agents Against Evolving Security Threats.
Conference on Language Modeling (COLM),
2025.
DoomArena: A framework for Testing AI Agents Against Evolving Security Threats.
Workshop at the International Conference of Machine Learning (ICML),
2025.
Silent Sabotage: Injecting Backdoors into AI Agents Through Fine-Tuning.
Workshop at the International Conference of Machine Learning (ICML),
2025.
StarVector: Generating Scalable Vector Graphics Code from Images and Text.
Computer Vision and Pattern Recognition (CVPR),
2025.
AlignVLM: Bridging Vision and Language Latent Spaces for Multimodal Understanding.
Workshop at the International Conference of Learning Representation (ICLR),
2025.
BigDocs: An Open and Permissively-Licensed Dataset for Training Multimodal Models on Document and Code Tasks.
International Conference of Learning Representations (ICLR),
2025.
InsightBench: Evaluating Business Analytics Agents Through Multi-Step Insight Generation.
International Conference of Learning Representations (ICLR),
2025.
LitLLMs, LLMs for Literature Review: Are We There Yet?.
Transactions on Machine Learning Research (TMLR),
2025.
StarVector: Generating Scalable Vector Graphics Code from Images and Text.
AAAI Demos,
2025.
BigDocs: A Permissively-Licensed Dataset for Training Vision-Language Models on Document and Code Tasks.
Workshop at the Neural Information Processing Systems (NeurIPS),
2024.