I’m Arya Hariharan, a final-year Computer Science student at R.V. College of Engineering, Bengaluru (CGPA: 9.85/10). My work sits at the intersection of machine learning, AI safety, and NLP, with a focus on building systems that are both capable and reliable.
Research Interests
I’m broadly interested in:
- AI Safety & Alignment — mechanistic interpretability, robustness of learned representations, and out-of-distribution generalization
- Medical AI — benchmarking and post-training of reasoning models for clinical tasks
- NLP for low-resource settings — bridging language barriers in education and legal research
Current Work
I’m currently part of two ongoing research efforts:
-
AI Safety Camp (AISC) — Working with an 8-member team led by Adrians Skapers (PhD, University of Manchester) on improving the generalizability of linear probes to out-of-distribution datasets, exploring ensembling, invariant risk minimization, and PCA-based techniques.
-
MedARC — Contributing to Sophont and MedARC’s open-source initiative on RL post-training and benchmarking of reasoning models for medical AI. I’m an author on Medmarks, a new LLM benchmark suite for medical tasks, currently under review at ICML.
Industry Experience
Microsoft India (R&D) — Software Engineering Intern (May–June 2025)
I worked on frontend features for Microsoft Teams Search, improving query responsiveness and click-through flows across 170M+ monthly user interactions. This involved optimizing React/TypeScript components to improve task completion rates at scale.
HPCC Systems, LexisNexis Risk Solutions — Project Intern (December 2023–October 2024)
I built an NLP-driven legal research assistant that extracts key phrases from case abstracts and retrieves relevant precedents using Named Entity Recognition and spaCy — achieving 15% higher precision over baseline techniques. The retrieval and ranking system was deployed on HPCC Systems using ECL, achieving sub-second responses on distributed data. This work won Best Poster – Use Case at the 2024 HPCC Systems Technical Poster Competition.
Academic Highlights
- CGPA: 9.85/10 (up to VI Semester), R.V. College of Engineering
- 99% in CBSE Grade 12 Board Examinations; State Rank 205 in KCET
- Certificate of Merit issued by CBSE for exemplary academic performance
- Best Outgoing Student Girl, 2022 Batch, Chrysalis High Kadugodi
Publications
I’ve published work across NLP, control systems, and computational biology. See the full list on my Papers page.
- 2026 Cohort — Centre for Long-Term Risk’s Fundamentals Program
- Winter Cohort — AI Safety Collab 2026 by ENAIS
- Member & Alumnus — Junior Academy & Thousand Girls Thousand Futures, New York Academy of Sciences
- Active member of the Frequency Club (technical) and Rotaract Club at RVCE
Feel free to reach out via email or connect on LinkedIn — I’m always happy to talk research.