AI & Software
NLP and LLM tools, from model training to the API behind them. Python, FastAPI, PyTorch, ChromaDB.
Third-year CSE · JNTUH (UCESTH)
AI and software engineering for practical problems.
I’m Shriya, a Computer Science student at JNTUH focused on applied AI, backend engineering, and workflow automation. I turn complex, document-heavy processes into clear, useful tools.
Open to AI and software engineering internships · Class of 2028
What I do
Three ways I work.
Visual Design
Brand identity, posters and campaigns for university events, kept consistent across every format.
Team leadership
Leading cross-functional hackathon teams from problem framing to a clear demo.
Featured projects
The projects I'd walk you through first.
Indian News Comparator
A Chrome extension that extracts a news article, scores its political leaning on a −5 to +5 scale, and compares it with related Left, Center, and Right coverage.
Version 1 combined LLM analysis with article retrieval. Version 2 uses an IndicBERT and Liquid Neural Network classifier, topic-filtered ChromaDB search, and Groq to explain differences in framing.
65.4% accuracy on 3 classes (33% baseline), trained on 2,594 labelled articles.
What I built
- The model training pipeline and bias classifier
- The FastAPI backend and API design
- Semantic retrieval with ChromaDB and topic-anchor filtering
- LLM-generated explanations with Groq
- Docker deployment and most of the Chrome extension
NeuralPath IISc Hackathon
Compares a résumé with a job description, identifies skill gaps, and builds a personalised day-by-day learning path with Llama 3.2 running locally through Ollama.
What I built
- The FastAPI backend
- PDF résumé parsing and skill-gap analysis
- Local Llama 3.2 through Ollama
- The adaptive learning-path algorithm
- Skill mapping to the O*NET taxonomy
First prototype
Media Bias Extension
The first version of Indian News Comparator. Groq classifies article bias and extracts entities; ChromaDB finds related coverage first, with NewsAPI and The Guardian as live fallbacks for Left, Center, and Right perspectives.
View repository →SprintMind Microsoft HackwithHyderabad 3.0
A team delivery intelligence tool that brings meeting and standup commitments together with Jira updates and GitHub pull request and CI events. Deterministic rules flag delivery risks with traceable evidence; Hindsight keeps dated team memory, and the LLM explains verified evidence.
View repository →Creative work
Selected design work.
Selected visual identity and campaign work for university events and student organizations, created with Affinity Suite and Canva.
Festival identity
Eraya
Design lead for a multi-college fest with 3,000+ attendees. Coordinated a large multi-team group and led the full event identity.
Brand & campaigns
E-Cell, JNTUH
Visual identity and social campaigns for university-wide startup events, supporting consistent communication across event materials and channels.
Campus community
Data Science Club & event support
Design head for Data Science Club events, plus volunteer coordination for the Global Alumni Meet and fests welcoming 1,000+ attendees.
Hackathons and competitions
Built under pressure.
I lead teams through fast-paced competitions, turning open-ended challenges into working prototypes and clear demos.
| Event | What I built | Role |
|---|---|---|
| Smart India Hackathon 2026 | TraceNet (main entry), QuantaDx (hybrid quantum-AI disease detection, backup idea) | Team lead and presenter |
| Amazon ML Challenge 2026 | Entity-resolution pipeline with LightGBM over millions of business records, 72 hours, team of 4 | Team lead |
| Microsoft HackwithHyderabad 3.0 | SprintMind: a scrum task agent that retains team context across sprints using Hindsight agent memory | Team lead |
| IISc Hackathon | NeuralPath | Team lead; backend and AI pipeline |
| Emerge 2.0, IIT Hyderabad | Entrepreneurship competition | Quarter-finalist |
| TCS Hackathon (AI track) | RetailVerse AI: a retail decision twin where an AI copilot's promotion edits are checked against inventory, margin and budget, then applied only after human approval | Team lead; frontend and model training (backend in progress) |
Amazon ML Challenge 2026
Entity resolution at scale
Built with a team of four in a 72-hour challenge: a LightGBM pipeline to match business records across millions of entries.
Smart India Hackathon 2026
TraceNet & QuantaDx
Led the team and presented TraceNet, with QuantaDx as the backup hybrid quantum-AI disease-detection concept.
TCS Hackathon · AI track
RetailVerse AI
A retail decision twin that checks an AI copilot's promotion edits against inventory, margin and budget before human approval. Team lead; built the frontend and trained the model, with backend work in progress.
Experiments
Smaller builds and things I made to learn.
Campus Loop
A campus hub for academic resources, viva preparation, placement stories and opportunities, anonymous student Q&A, events, and searchable campus knowledge.
Repository →ContextBridge
A local-first CLI and desktop app that records project decisions, commits, code changes, and test runs, then creates a concise handoff for the next AI coding agent. It works offline without an LLM.
Repository →NotionMake
A desktop app for visually building Notion pages and databases from templates or blocks, previewing them, and publishing to a workspace. Its optional AI assistant generates pages locally on your device.
Repository →Linear Regression From Scratch
A Python linear-regression model built from scratch to predict startup revenue from marketing spend, calculate MSE, and plot its predictions without an ML framework.
Repository →Railway Reservation System
A full-stack train-booking app with route search, multi-passenger reservations, payments, automatic seat allocation, and waitlist promotion after cancellations. I designed the Oracle SQL schema, constraints, and normalization.
Repository →Applied AI, dependable software, thoughtful design.
I’m a third-year B.Tech student in Computer Science at JNTUH’s University College of Engineering, Science and Technology, Hyderabad. My work spans NLP models, retrieval pipelines, LLM agents, and the FastAPI backends that support them. I focus on turning ambiguous requirements into practical, explainable tools.
Outside the code, I work on visual identity for campus events, lead hackathon teams, and present technical ideas to non-technical audiences. I also enjoy running models locally to keep user data on-device.
AI & Data
Build
Design
Working style
Experience and service
Selected campus contributions.
-
Club fest
Design Lead, Eraya
Led design for the whole fest (3,000+ attendees), coordinating a large multi-team group.
-
Ongoing
Design Head, E-Cell · JNTUH
Lead visual identity and design campaigns for campus startup events.
-
Ongoing
Design Lead, Data Science Club · JNTUH
Lead visual design for club events and campaigns.
-
Volunteer
JNTUH UCESTH
Coordinated the Global Alumni Meet and on-ground logistics for fests with 1,000+ attendees.
Education
B.Tech, Computer Science and Engineering
JNTUH University College of Engineering, Science and Technology, Hyderabad. 3rd year, expected 2028. CGPA 9.295.
Class XII: Narayana Jr. College, 96.2% (2024), JEE Main 98.7 percentile. Class X: D.A.V. Public School, 96.17% (2022).
Workshops: Agentic AI, AI fundamentals, AI and Data Science Intensive (2024).
Contact
Let's build something.
I'm looking for AI and software engineering internships, and I'm always up for a conversation about a project, a design idea or something you're building.