Hi, I'm Dhruv Jain.

I am a Research Engineer at Straiker, working on AI safety & security. My work focuses on robustness, adversarial evaluation, and secure deployment of agentic systems in real-world environments.

Previously, I was a Research Engineer at Ola Krutrim, where I worked on post-training, evaluation, and deployment of large-scale language and speech models for production AI assistants. My work there focused on improving safety, retrieval and search, and tool orchestration in real-time systems, with an emphasis on multilingual and Indic AI. I built VoiceAgentBench, the first multilingual benchmark for evaluating voice assistants on tool use, multi-turn planning, and adversarial safety.

Earlier, I was a Research Intern at MIDAS Lab, IIIT Delhi (with Prof. Rajiv Ratn Shah) on improving STEM reasoning in LLMs, and at NTU Speech Lab (with Prof. Chng Eng Siong) on ASR post-processing and domain adaptation. I completed my Bachelor’s degree in Electronics Engineering at IIT (BHU), Varanasi. For more details, check out my resume.

You can reach me via email or LinkedIn.

News

• Mar 2026. Joined Straiker as Research Engineer to work on AI Security.
• Feb 2026. Open-sourced VoiceAgentBench with the Hugging Face dataset.
• Nov 2025. Improving Physics Reasoning in LLMs using Mixture of Refinement Agents accepted at AAAI 2026 TrustAgent Workshop.
• Jun 2025. Joined Ola Krutrim as Research Engineer.
• Jul 2024. Appointed General Secretary at Science and Technology Council (IIT BHU).
• May 2024. Joined NTU Speech Lab as Research Intern.
• Mar 2024. Joined MIDAS Lab, IIIT Delhi as Research Intern.
• Feb 2024. Released arXiv preprint of our work SwissNYF: Tool-Grounded LLM Agents for Black-Box Settings.
• Jul 2023. Appointed Joint Secretary at Robotics Club (RoboReG), IIT BHU.

Selected Publications

• VoiceAgentBench: Are Voice Assistants Ready for Agentic Tasks?

  Dhruv Jain*, Harshit Shukla*, Gautam Rajeev, Ashish Kulkarni, Chandra Khatri, Shubham Agarwal

• Improving Physics Reasoning in Large Language Models Using Mixture of Refinement Agents

  Raj Jaiswal*, Dhruv Jain*, Harsh Parimal Popat, Avinash Anand, Abhishek Dharmadhikari, Atharva Marathe, Rajiv Ratn Shah

• SwissNYF: Tool Grounded LLM Agents for Black Box Setting

  Somnath Sendhil Kumar, Dhruv Jain, Eshaan Agarwal, Raunak Pandey

See full publications page.

Projects

• VoiceAgentBench

  Multilingual benchmark for speech-based agents with 6k+ spoken queries across English, Hindi, and five Indic languages.

• Self-Correcting RL for Physics

  Two-stage RL pipeline with step-level error feedback to improve reasoning reliability in small language models.

• Dynamic Multi-Agent RAG System

  Interleaved retrieval and reasoning for long legal/financial documents, with robust multi-turn tool-calling.

See full projects page.