First-Year Master's Student @ MBZUAI

Kapil Rajesh Kavitha

Graduate NLP researcher focused on cultural alignment and fairness, and evaluation of multilingual LLMs. I build applied ML systems with an emphasis on real-world deployment, and I love pursuing projects that match both my research and non-academic interests.

Download Resume View Projects
Phone: +971 542217325
Portrait of Kapil Rajesh Kavitha

Education

Academic background in NLP and computer science.

Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI)

M.Sc in Natural Language Processing – Aug 2025 – Present

CGPA: 3.75 / 4

International Institute of Information Technology (IIIT Hyderabad)

B.Tech in CSE with Honors – Nov 2021 – Jul 2025

CGPA: 8.24 / 10

Experience

Research and industry experience across NLP and applied ML.

Graduate Researcher – MBZUAI

Oct 2025 – Present – LLMs, vLLM, Evaluation, Bias & Safety

  • Benchmarked 20 multilingual LLMs on Indic-Bias for fairness across English and Indian scripts.
  • Analyzed bias magnitude, stereotype rates, and refusal-to-answer behavior.

Data Science Intern – Athlyte Inc

May 2024 – Jan 2025 – LLMs, Python, MongoDB, AWS

  • Built LLM-powered tweet classification, entity extraction, and structured query generation.
  • Developed YOLO + AWS Textract table extraction pipeline with ~80% accuracy.

Researcher (Honors) – MT-NLP Lab, IIIT Hyderabad

May 2023 – May 2025 – Transformers, PyTorch, Hugging Face, LLMs

  • Developed encoder-decoder models for code-mixed MT with finetuning and evaluation frameworks.
  • Studied Indic LLMs for English-Hindi code-mixing using LoRA-based pretraining.

Teaching Assistant – IIIT Hyderabad

Jul 2023 – Dec 2024 – Advanced NLP, Intro to Software Systems

  • Created assignments, ran tutorials, and graded exams for 200+ students.

Projects

Selected projects in NLP, LLM systems, and applied research tooling.

Football-QA

Python, SQLite, LLMs, Tool Calling, Gradio

LLM-powered QA agent for football queries over a local SQLite database.

View on GitHub

Early-Exit Transformers for Adaptive Computation

Transformers, PyTorch, Efficient Inference

Implemented early-exit blocks with confidence-based gating and shared KV-cache.

View on GitHub

LLM Debater

LangGraph, Multi-Agent Systems, Gradio

Multi-agent debate system with state tracking and an LLM judge agent.

View on GitHub

UpDawg

Software Engineering, Design Patterns, Python, Flask, LLMs

Personal assistant prototype aggregating Slack/Outlook data with LLM summaries.

View on GitHub

AS2 – Answer Sentence Selection

Python, PyTorch, Transformers

Two-step fine-tuning pipeline for answer selection with near-SOTA results.

View on GitHub

Publications

Selected publications from 2024.

Benchmarks Underestimate the Readiness of Multilingual Dialogue Agents

Lee, A. H., et al. – 2024

Bridging Laughter Across Languages: Generation of Hindi-English Code-Mixed Puns

Asapu, L., et al. – 2024

Technical Skills

Tools and coursework that support my research and engineering work.

Programming

  • Python, C++, JavaScript, SQL
  • PyTorch, Transformers, Hugging Face, vLLM, WandB

Data & Infrastructure

  • Docker, MongoDB, Pandas, NumPy
  • Advanced NLP, Deep Learning, ML, Probabilistic & Statistical Inference

Resume

Download a copy of my latest resume.

Resume PDF

Download Resume

Contact

Interested in collaboration or research? Let’s connect.

Get in touch

Currently

  • Graduate researcher at MBZUAI
  • Open to collaborations in cultural alignment, fairness, and LLM evaluation