cv

Software Development Engineer Intern — Alexa Audio

Amazon

• Built A3LS, an accent-audio testing service generating TTS across 13 accents and scoring per-accent ASR degradation (WER/AIER), via a Step Functions pipeline (Java Lambda, Coral/API Gateway) that fans out over tens of thousands of (utterance x accent) items per run.
• Shipped a React/Cloudscape analytics dashboard over a CloudWatch metrics API, surfacing per-accent WER/AIER, latency percentiles, and GPU/concurrency, with skeleton loading and caching for instant loads.
• Authored HLD/LLD design docs and drove cross-team design reviews to integrate an accent dimension for injecting synthetic audio into production ASR test runs.
• Delivered across 80+ code reviews spanning 4 Brazil packages (Smithy, Java, CDK, React), owning CI/CD via Pipelines/Apollo.

Software Engineering Intern

Nomura

React, Java Spring Boot, SQL, Jenkins, Agile.

• Worked on an internal budgeting and financial planning application used across multiple teams.
• Resolved over 40 JIRA tickets in 6 months, focusing on performance improvements, bug fixes, and user experience.
• Delivered features using React on the frontend and Spring Boot for backend services.
• Collaborated within an Agile team, using Jenkins for CI/CD and SQL for data-layer operations.

Backend Intern

AIAssistant Startup

A startup focused on voice and agentic AI solutions. FastAPI, Python, Postman.

• Developed backend solutions for multiple clients using Twilio, Deepgram, and ElevenLabs.
• Built prototypes for client demos, including call summarization, healthcare assistants, and lead-generation voice systems.

Data and GenAI Intern

Ernst and Young

• Worked on business and technical solutions for Fortune 100 clients driven by Generative AI.
• Developed a chatbot with a RAG pipeline over 10,000+ documents, improving search efficiency by 200% via metadata filters that narrowed the search space from 10,000 documents to 50.

Research Intern — Prompting and LLMs

IIT Kharagpur

• Participated in red-teaming to identify and analyze safety limitations and vulnerabilities of LLMs.
• Implemented several chain-of-thought strategies to strengthen LLM robustness and ethical application, with custom Python scripts for automated output evaluation.
• Evaluated jailbreak robustness of ChatGPT, Gemini, Nous-Hermes, Mixtral Instruct, Cohere Coral and other LLMs using custom multilingual prompts, cross-evaluating outcomes with multiple HuggingFace sentiment models.

Research Intern — Natural Language Processing

IIT Indore

• Extracted critical information from disaster-related tweets using advanced text summarization, reducing a dataset of 3,000 tweets to the 40 most crucial.
• Applied large language models and integer linear programming to improve summarization accuracy and efficiency.
• Engineered a flood detection project applying segmentation techniques to improve classification outcomes.

Machine Learning Intern

RCI — Defence Research and Development Organisation (DRDO)

• Used Splunk and Kafka to improve the system’s security measures.
• Performed log analysis and anomaly detection on over one million logs, reducing major outliers to a few hundred; the proof of concept was showcased to senior scientists for possible production deployment.
• Applied One-Class SVM, KMeans, and Isolation Forests to improve detection accuracy.
• Developed a dynamic web application with HTML, CSS, JavaScript, Bootstrap, and Flask.

Artificial Intelligence Intern

Jio Platforms Limited

• Worked on “Improving Operations with AI”, focused on improving business operations using advanced AI.
• Designed and deployed recommendation algorithms, including Apriori, to enrich decision-making.
• Built a web application presenting the implemented features to key stakeholders.

Master of Science

University of North Carolina at Chapel Hill

Computer Science

• Focused on NLP, AI, and research-driven development.
• Teaching Assistant for DATA 541 — Natural Hazards and Financial Risk (Fall 2026).
• Teaching Assistant for DATA 520 — Research Methods for Socially Responsible AI (Fall 2025).
• ASPSA tutor (Spring 2026) for MATH 110, DATA 110, and COMP 110, supporting 40+ student-athletes.
• Researcher in the Society-Centered AI Lab (SAIL) with Dr. Neil Gaikwad on Responsible AI.
• Researching with Prof. Snigdha Chaturvedi on LLM Alignment.

Honors

Indian Institute of Information Technology Pune

Machine Learning

• Ranked 1st in batch.

Bachelor of Technology

Indian Institute of Information Technology Pune

Computer Science Engineering

• Consistently ranked among the top 3 out of 175 students.
• Sole student to score a 10/10 SGPA in the 7th semester.

SHARD: Safe and Helpful Alignment via Self-Reframing Distillation

EMNLP 2026 Findings

Large language models often struggle with sensitive prompts — refusing outright, providing generic safety boilerplate, or failing to address legitimate informational needs that can be answered safely. SHARD is a self-reframing distillation method to improve safe-helpfulness.

Compounding Disadvantage: Auditing Intersectional Bias in LLM-Generated Explanations Across Indian and American STEM Education

ACM Conference on Fairness, Accountability, and Transparency (FAccT) 2026

Audits intersectional bias in AI-generated explanations across Indian and American STEM educational contexts, showing how disadvantages compound across language, caste, and social context rather than simply adding up.