Hi, I'm Aravind Cheruvu.

• Resource-Aware Agentic AI / Efficient Inference: Research inference-time methods that make LLM reasoning and AI agents more compute-efficient while leaving model weights unchanged.
• Reasoning Budget Control: Study how reliably reasoning models can be held to a thinking-token budget, comparing prompting, decoding-time truncation, and activation steering.
• Confidence-Calibrated Early Stopping: Develop lightweight controllers that use calibrated model confidence and answer stability to decide when a multi-step agent can stop without losing accuracy.

• Generative AI / LLMs / Health Applications: Designed and developed Generative AI applications for digital health and wellness, delivering adaptive coaching, personalized recommendations, and context-aware health insights.
• Backend Prototyping / RAG / Cloud Integration: Developed scalable backend pipelines and RESTful APIs to process multi-modal health data from smartphones and wearables, enabling RAG-based AI assistants that provide real-time, evidence-informed guidance.
• Cross-Functional Research / Explainable AI / Clinical Collaboration: Partnered with AI scientists, clinicians, and human-factors researchers to co-innovate digital health solutions. Implemented AI-driven insights and explainable visualizations translating wearable and time-series analytics into actionable wellness feedback. Supported pilot studies to validate GenAI-based health coaching and conversational frameworks.

• Research focus: Security and Generative AI under the supervision of Dr. Danfeng Yao.
• Projects: Toxicity injection attacks on chatbots, deepfake detection, and LLM safety alignment.
• Achievements: Published papers at top-tier conferences (ACSAC 2023, IEEE S&P 2024, CODASPY 2026).

• Certified Oracle HCM Cloud transformation consultant with 4.5 years of demonstrated techno-functional expertise specialized in capturing business use cases, understanding requirements, and performing fit-gap analysis to design scalable 50+ Technical RICEF objects and business process solutions.
• Strategized and executed Payroll Parallel/Reconciliation testing cycles for 5 successful client implementations to uncover system implementation defects, understand financial and business process impacts of Go-live and Post Production, and recommend mitigation strategies.

Payroll Reconciliation:

• Led planning and execution of Payroll Compare cycles for multiple clients to perform trend analysis of $MM employee payrolls, test data integrity, and understand Go-Live and Post-Production impacts of Payroll, Benefits, Time (TL), Absence, and Compensation systems using SQL and Excel analytics.
• Developed and streamlined a Payroll Compare Analysis and Executive Reporting Tool which extracts and uses HR and Payroll run data between Legacy and simulated Test systems to produce Payroll Compare reports and Executive Dashboards for extensive payroll data analysis.

Payroll RICEF:

• Supported Payroll BR100 configurations and coordinated and executed SIT and UAT testing cycles.

Technical Developer:

• Worked as a technical team member implementing key out-of-the-box integrations using HCM Extracts to Kronos and Benefits systems, BI Publisher reports using eText and RTF templates, and developed Payroll Fast Formulas.

• Trained in E-Business Suite, Oracle Business Intelligence EE, and Oracle Data Integrator tools.

• CGPA: 3.77
• Advisor: Dr. Danfeng Yao
• Research focus: Security and Generative AI
• Status: Ph.D. completed - dissertation defended (June 2026)
• Dissertation: "Defending the Dialog: Characterizing and Mitigating Harmful Fine-Tuning Attacks in Conversational AI"

• CGPA: 8.51 / 10.0
• Achievement: Gold Medal for best outgoing student from the Department of I.T.
• Specialization: Information Technology with focus on Data Mining and Network Security