I'm a third-year CSE student at GITAM University in Visakhapatnam (CGPA 9.17), interested in data engineering and backend work. I would rather build a pipeline that copes with bad input than tune a model for another half a point of accuracy.
- Software engineering intern at Motherson Technology Services (AIML team). I built the chat-history storage for LegalLens, an AI legal-research platform: a FastAPI module with 9 endpoints, SQLite for metadata and Azure Blob Storage for message payloads.
- IBM Data Engineering Professional Certificate (16 courses).
- President of IEEE-CS GITAM Vizag, and Domain Lead for Open Source and DevX in the GitHub Community at GITAM Vizag.
- Currently learning Airflow, star-schema design, incremental ETL, and more FastAPI and PostgreSQL.
I'm open to data engineering and backend internships.
- Languages: Python, C, C++, Java, SQL, TypeScript, JavaScript
- Data: Pandas, NumPy, Airflow, Kafka, Spark and Hadoop (from the IBM course), Cognos and Looker
- Databases: PostgreSQL, MySQL, MongoDB, SQLite
- Backend: FastAPI, Node.js, Express, React, Mongoose, JWT
- ML: scikit-learn, XGBoost, PyTorch, Sentence Transformers, SHAP, NetworkX
- Tools: Git, GitHub Actions, Docker, Linux, Playwright, Make, GDB, Valgrind, Azure Blob Storage, Vercel
Python, XGBoost, SHAP, FastAPI, React, NetworkX. Built for a Smart India Hackathon problem statement from ITBP / Ministry of Home Affairs (SIH-1775).
- Two XGBoost classifiers trained on about 101,000 labeled accounts: 90.04% accuracy for X and 96.87% for Meta platforms.
- A noisy tweet-level dataset was holding X accuracy at 67%. Removing it brought the model to 90.04%.
- SHAP explanations shown as plain-English risk factors, a NetworkX graph for coordinated accounts, and PDF case-file export.
- Playwright scraping that never stores credentials.
Python, Pandas, NumPy, BeautifulSoup, SQLite. Scrapes market capitalization data for the largest banks, converts currencies with vectorized operations, and loads the result into SQLite and CSV, with timestamped logging at each stage.
Node.js, Express, PostgreSQL, JWT, bcrypt. A clinic app where students book doctors and doctors record diagnoses and prescriptions. Built at a 24-hour university hackathon. The repo has 29 integration checks.
TypeScript, React, Express, MongoDB, Gemini. A road-trip planner with maps, nearby stops and a chat assistant. Started for Smart India Hackathon 2025. It is still in development.
- mealbridge: MERN app that connects kitchens with surplus food to NGOs (live)
- astra-3d-music-universe: WebGL scene built from Spotify listening history (live)
- sih_149: secure data erasure and file recovery tool with a hash-chained audit log
- sluice, terminal-arcade and open-source-launchpad: practice repositories I set up for first-time contributors at GITAM
- Smart India Hackathon 2026: internal-round qualifier, fake account detection engine (SIH-1775)
- Smart India Hackathon 2025: internal-round qualifier, TripSync
- 24-hour university hackathon: KAPS MedTech backend
The contribution graph and the snake are regenerated daily by the workflows in .github/workflows.