Nishitha Khasnavis

About me

Hello! My name is Nishitha and I'm a computer science student at the University of Virginia. I have a passion for building purposeful and scalable software solutions that have a meaningful impact.

Through my internships and research experience, I’ve built full-stack applications, deployed cloud-native systems, developed ML pipelines and fine-tuned large language models . I am also actively involved in UVA’s Girls Who Code and the Society of Women Engineers - communities that have shaped my interest in ethical, inclusive innovation.

I am excited for the opportunity to apply these skills in mission-driven environments, where I get to contribute to developments that make people's lives better. As I continue learning and growing, I look forward to collaborating on purposeful projects that challenge me and push the boundaries of what’s possible.

Feel free to explore this website to learn more about my work — or contact me here! You can also email me at nishitha.khasnavis@gmail.com

Software Engineer Intern @ VocsAI

(March 2025 - July 2025)

As a Software Engineering Intern at Vocs AI, I led the backend development of a cloud-native voice synthesis platform serving over 25,000 users. The platform’s core challenge was scaling real-time audio generation while keeping latency minimal and performance consistent — an ideal opportunity to bridge systems design with creative AI applications.

My primary focus was designing a robust backend architecture capable of handling thousands of audio requests concurrently. I engineered PostgreSQL and MongoDB pipelines for structured and unstructured data, integrating Redis caching to improve lookup efficiency. This optimization reduced retrieval latency by nearly 70%, ensuring the system could support fast playback and generation without lags.

I built REST and GraphQL APIs using FastAPI, optimizing for asynchronous execution and load balancing. These APIs managed over 200 requests per second, coordinating voice generation, playback, and metadata management. Each endpoint was secured and logged for monitoring through production-level observability tools.

For deployment, I containerized services using Docker and orchestrated infrastructure on AWS EC2, managing assets through S3 and CloudFront. I automated updates with CI/CD pipelines, achieving zero-downtime deployments across environments — a milestone that strengthened the platform’s reliability.

This experience honed my expertise in backend scalability, distributed systems, and API design, but also taught me the nuance of maintaining engineering precision in fast-moving production environments. It reaffirmed my passion for building systems that merge technical robustness with real-world usability — where milliseconds of latency can define the quality of user experience.

To manage data effectively, I designed relational schemas with SQL for user and session data, and used MongoDB for optimized storage of audio metadata in a document-based structure. I also implemented and deployed RESTful APIs to improve data flow across services, which reduced frontend latency by approximately 15%, resulting in a smoother user experience.

On the infrastructure side, I deployed the application to AWS EC2, managed static content with S3, and gained hands-on experience in orchestrating a cloud-native architecture that scales efficiently. This experience gave me practical exposure to cross-service communication, cloud deployment workflows, and performance monitoring in production environments.

Through this project, I strengthened my understanding of end-to-end web development, API design, cloud infrastructure, and building real-world systems that are both scalable and user-centric.

Machine Learning Researcher @ Aikyam Lab (UVA)

(May 2025 - Present)

At the Aikyam Lab at the University of Virginia, I’m researching how we can make recommender systems not only smarter, but more responsible. My project focuses on machine unlearning — figuring out how large language model–based recommenders can “forget” user data when privacy laws or ethical standards require it, without sacrificing accuracy.

To support this, I built a PySpark + Docker pipeline that processes more than 10 million user-item interactions, running across Kubernetes clusters on UVA’s high-performance computing system. The pipeline structures and cleans the data so our models can train faster and more reliably. Using MLflow, I automated tracking across experiments, which made our retraining workflow reproducible and transparent.

I fine-tuned the LLaMA-3.2-1B model with LoRA, a method that lets us adapt massive language models using only a fraction of the memory. The optimized version improved precision by 30%, measured through NDCG@5 and MRR@10. I also experimented with semantic search and BM25 reranking, recreating the kind of retrieval dynamics users actually experience in a product recommendation setting.

This work has been both technically demanding and conceptually eye-opening. It’s taught me that machine learning isn’t just about better predictions — it’s about earning trust through transparency and control. Working alongside Professor Chirag Agarwal and Professor Sam Levy has shown me what it means to build AI that’s not only high-performing, but also human-aligned and ethically sound.

Headstarter Fellowship

(June 2024 - August 2024)

Through the Headstarter Fellowship, I collaborated with engineers across time zones to build a real-time collaborative code editor and a set of AI-powered learning tools. Our goal was to make remote coding and study sessions feel as natural as sitting side-by-side.

I helped design the frontend using React and the Monaco Editor, then wired everything together with WebSockets so edits appeared instantly for all participants. Behind the scenes, I worked on an AI flashcard generator using LangChain and Databricks that reduced manual prep work by 60%, turning long lecture transcripts into interactive study material.

We deployed the platform on Microsoft Azure, adding auto-scaling and load balancing so the app could handle spikes in usage without slowing down. Throughout the process, I learned how to debug distributed systems in real time, balance user experience with server cost, and ship features fast without breaking what already worked.

What I loved most was how every design choice had a human consequence: every extra millisecond of latency changed how people collaborated, every prompt adjustment made the AI more or less useful. The fellowship reinforced my belief that engineering is ultimately about people — understanding their needs, constraints, and rhythms, then using technology to quietly make their work easier.

Student Researcher @ Prof. David Evans (UVA)

(January 2024 - June 2024)

In recent years, Large Language Models (LLMs) have rapidly advanced and become integral to a wide range of applications - from chatbots and virtual assistants to content generation tools and personalized search engines. However, these systems often operate as black boxes and can unintentionally leak sensitive information, especially when trained on user-generated or proprietary data. As part of my research with Professor David Evans, I am exploring ways to systematically identify, simulate, and mitigate these privacy risks through automated information disclosure audits.

My work involves developing a novel methodology that enables auditing LLM-based systems for potential information leakage. I focus on an approach called auditing-by-parity, where we simulate an adversary who queries the system with crafted inputs designed to elicit sensitive responses. To conduct these audits, I utilize OpenAI APIs in combination with advanced NLP frameworks like Hugging Face Transformers and PyTorch. This setup allows me to reproduce real-world attack scenarios and assess how much confidential or memorized information the model may expose.

One of the key components of this research has been the development of a robust and modular pipeline for data preprocessing, model training, and evaluation. This pipeline automates the process of crafting adversarial prompts, querying LLMs, and evaluating the responses.

This research sits at the intersection of machine learning, security, and privacy - areas that I am deeply passionate about. It requires not only a strong understanding of model deep learning workflows but also creative thinking to simulate threats and devise countermeasures.

The broader goal of this work is to help developers, researchers, and organizations deploy LLMs more responsibly - with greater transparency, auditability, and trust. As LLMs become ubiquitous, ensuring that they don’t leak sensitive training data is not just a technical issue but a societal one. Through this project, I hope to contribute to building safer and more privacy-aware AI systems.

CoveStack

CoveStack is a personal project born out of a common frustration I face as a developer - constantly switching between disconnected tools like GitHub, Slack, Notion, VSCode, and shared docs during team-based projects. This fragmentation often breaks flow, slows collaboration, and makes it harder to focus on building. I realized this isn’t just my problem - it's something many engineers experience daily.

To solve this, I started building CoveStack, a cloud-native collaboration platform that brings together code sharing, task management, and real-time communication into a single, unified workspace. The vision is to streamline workflows and create a tool that lets teams focus on solving problems and significantly improve productivity.

I led full-stack development using a React + TypeScript frontend and a modular FastAPI + Node.js + Go backend. The frontend architecture emphasized component reusability and speed, resulting in 40% faster load times across devices. I integrated OpenAI GPT APIs to enable semantic search — allowing users to query workspace history in natural language and surface contextually relevant notes and documents instantly.

The backend was designed for performance: distributed microservices, asynchronous API calls, and caching mechanisms allowed the system to handle 500+ concurrent requests seamlessly. I enforced strict API testing using Pytest, Jest, and Postman, cutting debugging time by 30% and maintaining production stability.

I also focused heavily on user experience — from responsive design to intuitive onboarding — ensuring that the platform felt effortless even as it managed complex state synchronization.

CoveStack reflects my interest in building systems that are functional, scalable, and developer-friendly. It aligns closely with my passion for developing innovative solutions that not only improve individual productivity but also foster a stronger sense of collaboration and shared purpose within developer communities.

NeuroVision

Under the mentorship of Professor Manoj Patel, I worked on NeuroVision, a deep learning system for detecting epileptic seizures in lab mice. The goal was to replace manual observation with automated detection through computer vision and neural networks, improving the speed and consistency of neuroscience research.

I developed a Convolutional Neural Network (CNN) using PyTorch and OpenCV, training it on video data captured from experimental trials. The preprocessing pipeline extracted motion patterns and segmented activity frames, reducing visual noise and improving clarity. I engineered modular data workflows backed by PostgreSQL, streamlining access and analysis across thousands of video samples.

Through iterative tuning, the model achieved a 20% improvement in baseline accuracy, successfully identifying subtle seizure onsets that human observation often missed. I visualized model outputs with Matplotlib and implemented logging, testing, and version control through GitHub to ensure reproducibility.

Beyond the technical work, NeuroVision showed me how AI can accelerate biomedical discovery — turning data into actionable insight. The project combined my interests in deep learning, data engineering, and computational neuroscience, reinforcing my belief in using AI to advance both science and human understanding.

UVABites

UVABites was born from a simple idea: reduce campus food waste while making free food events more accessible to students. Alongside a small team, I built a React + Firebase web app that connects students to real-time food availability across the University of Virginia.

I implemented Google OAuth authentication, push notifications, and Firebase ETL pipelines integrated with AWS S3, enabling instant updates as new food sources were added. These optimizations improved reliability and cut debugging time by 25%.

To manage scalability, I applied cloud design principles — leveraging Firestore’s real-time database and secure storage to support a growing base of over 20,000 students. Working in Agile sprints, I collaborated cross-functionally, using GitHub for version control and conducting iterative feature testing with peers.

Beyond the code, UVABites became a lesson in social impact engineering — proving that small technical innovations can create tangible community benefits. It strengthened my skills in frontend development, database design, and cloud deployment, while deepening my motivation to use technology for sustainability and civic good.

Technical Skills

Languages

Python, Java, C, Go, JavaScript, TypeScript, SQL, HTML/CSS, MATLAB, YAML, Shell (Bash)

Frameworks & Libraries

React, Angular, Node.js, FastAPI, Django, PyTorch, TensorFlow, Hugging Face Transformers, LangChain, MLflow, Unsloth, TRL

Data & Machine Learning

PySpark, Pandas, NumPy, OpenCV, ETL Pipelines, LLM Fine-Tuning, NLP, Recommender Systems, Retrieval-Augmented Generation

Cloud & Infrastructure

AWS (EC2, S3, CloudFront), Microsoft Azure, Kubernetes, Docker, HPC (Slurm), Redis, CI/CD (GitHub Actions, Pipelines), Linux

Databases

PostgreSQL, MongoDB, Redis

APIs & Integrations

REST & GraphQL APIs, OpenAI API, Stripe API, Groq API, OAuth2, JWT, RBAC

Collaboration & Tools

GitHub, Git, Figma, Adobe XD, Microsoft Teams, Slack, Microsoft Office Suite (Word, Excel), Agile Development

Let's Connect

• Linkedin

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';