ServiceNow AI Research

Luke Kumar

Luke Kumar

Applied Research Scientist

Agentic Harness & Defenses

Luke is an Applied Research Scientist at ServiceNow AI Research, researching guardrails and context-driven defense pipelines to secure enterprise AI agents against adversarial adaptive attacks. I also contribute to the Apriel family of foundation models — including Apriel-H1, a hybrid Transformer–Mamba architecture — with work spanning the full model lifecycle: pre-training, mid-training, SFT, distillation, and evaluation.

Previously at Microsoft and Amii, I built transformer-based B2C products and led applied ML initiatives in Health, Finance, and Energy, helping organizations adopt and deploy machine learning effectively. I bridge research and real-world impact — turning emerging model innovations into secure, high-impact AI systems.

Interests
  • Natural Language Processing
  • Large Language Models
  • AI Agents
  • Security

Publications

Using Scaling Laws for Data Source Utility Estimation in Domain-Specific Pre-Training. Conference on Language Modeling Workshops,  2025.

Paper Cite