Vineela Therala

I build Python data and LLM applications, and I write down what worked and what didn’t.

B.E. Computer Science and Data Science at MVSR Engineering College, Hyderabad. Graduating in 2028. Interested in AI/ML, AI agents, explainable AI and open source.

Projects GitHub LinkedIn Resume (PDF)

About

I am a Data Science student turning coursework into working software: data analysis in Python, small ML models, and apps that combine APIs with language models.

I label every project with what is finished and what is not, including a pull request that was closed and a model that reached about 60% accuracy. Writing that down is how I find what to learn next.

Vineela working on a laptop during an AI/ML mock-interview practice session

Projects

One team project that is live, one small model that taught me about data quality, and one deployed app.

News Navigator AI

Live Team project

Problem

Understanding one topic usually means reading many articles. News Navigator takes a topic and returns a short briefing, so a reader can decide what is worth reading in full.

What it does

The site collects recent articles on a topic and has a language model write a summary, key points, different perspectives, possible market impact and three follow-up questions.

My part

We built it as a team. I wrote the impact model and presented the project in our demo video.

What I measured and what I only estimated

Nothing has been measured with users. My impact model assumes a reader opens 8 articles at about 3 minutes each, against about 30 seconds with the tool. That is an assumption, not a result.

Limits and next steps

The briefing is written by a language model, so it can be wrong. The next improvements are showing each article’s title, source and link beside the briefing, limiting the model to the fetched articles, clearer error messages, and automated tests.

PythonFastAPIGroq APILlama 3.1Vercel

Open the live app

Data flow, from the architecture slide in our demo video. The video shows a Streamlit interface; the live version is hosted on Vercel.

Student Attention Classification

Learning project

An image classifier that tells “reading” from “using a phone”, built in Google Teachable Machine on 350+ images I collected in different conditions.

About 60% accuracy

Why it was low

Many of my images were blurred, filtered or unclear, so the model had little clean signal to learn from.

Collect cleaner, consistent images and learn how model evaluation, overfitting and data splits work before rebuilding it in Python.

Google Teachable MachineImage classification

Timetable Generator

Deployed

A web app that builds a timetable from scheduling rules. I made it for a hackathon. It is rule-based, not machine learning.

PythonPandasStreamlit

Open the app

Open source

Each contribution shows its real status, including the ones that were not merged.

gwlearn (PySAL), PR #45

gwlearn models need spatial geometry as an extra input, which breaks scikit-learn tools such as GridSearchCV. I proposed an adapter that carries geometry through scikit-learn’s metadata routing and wrote a demo script that tunes a geographically weighted regression. I also commented on another contributor’s pull request on the same topic.
View the pull request Read my write-up
Opened
StingrayExplorer (OpenAstronomy), PR #1

The dashboard failed to start with a ModuleNotFoundError. Python imports are case-sensitive, and an import in HomeContent.py did not match the real file name. I corrected the import and added a missing __init__.py to the utils folder. The maintainer closed the pull request in November 2025 without merging it. I also documented the installation errors I hit for other contributors.
View the pull request
Closed, not merged
GSoC 2025 proposal for OpenAstronomy

I wrote “Interactive Database for X-ray Observations” for the Stingray project. The GSoC dashboard ranks it #1 for the project, and it was not selected.

Not selected

Background

Education, credentials, what I am exploring, and what I do outside code.

Education

B.E. Computer Science and Data Science, Maturi Venkata Subba Rao Engineering College (MVSR), affiliated with Osmania University, Hyderabad. Expected 2028.

Certifications and programs

Cloud Virtual Internship, AICTE–EduSkills with AWS Academy curriculum. 8 weeks, April to June 2026. Completed.
ET-AI Hackathon 2026 (The Economic Times). Semi-finalist; certificate dated 7 May 2026.
Deloitte Australia Technology Job Simulation. Completed.
ISRO Geospatial Remote Sensing certification. Completed.
AICTE Virtual Internship, AI for India 2.0. Shopping-trends analysis with Pandas and Matplotlib.
Tata Imagination Challenge 2025. Participated in the Tata Quiz.
Oracle AI agents course. In progress.

Research interests

How explainability and fairness checks can be applied to vision AI models. This is at the idea stage: I am reading background material and have no experiments or results yet.

Leadership and events

NCC cadet and holder of the NCC B Certificate.
AWS Builder Center: Future Builders Connect, HITEX Exhibition Centre, Hyderabad, 17 September 2026. I took part in a hands-on workshop on building AI agents with Amazon Bedrock, plus cloud and agent-prompting activities and quizzes.
Took part in an AI/ML mock-interview practice session at MVSR.
Attending DevFest Hyderabad on 24 October 2026.