40.7440° N, 74.0324° W Hoboken, New Jersey

GyanasaiAdusumilli

Aspiring data scientist working on ML models

Master's student in Data Science at Stevens Institute of Technology. I like building machine learning projects and learning something new with each one.

View my work GitHub LinkedIn
Gyanasai Adusumilli

Graduating December 2026
Open to ML, AI and data science roles

PyTorchscikit-learnHugging FaceLangChainFAISSOpenCVGoogle Earth EngineFastAPIDockerPostgreSQLMySQLPython

Selected work

Some of the projects I've worked on during my studies.

Remote sensing

Water Body Detection

Pixel60, 45
NDWI0.00
ClassLand

Maps surface water from Sentinel-2 and Landsat-8 imagery. Five classical models and two index baselines go head to head with a U-Net CNN on a spatially split test set.

U-Net: 0.984 water IoU on Sentinel-2, versus 0.950 for the classic NDWI threshold. On 30 m Landsat-8, gradient boosting edged it out.

Try it: switch views and hover the tile to read each pixel's water index.

  • PyTorch
  • U-Net
  • scikit-learn
  • Earth Engine
  • Rasterio
Read case study Code
Learning to rank

Hotel Search Ranking

Ranks hotels in 400k real Expedia searches so the ones travellers book appear first. Six rule-based rankings and two pointwise models go up against LambdaMART, scored on future searches with NDCG and offline A/B tests.

Booked hotel in the top 5 for 58.8% of searches, versus 42.4% for the best rule. Statistically level with Expedia's own production ranking.

  • LightGBM
  • LambdaMART
  • Feature engineering
  • A/B testing
  • pandas
Code and results
NDCG@5 on 39,862 future searches
Random order0.159
Cheapest first0.205
Most booked first0.262
Pointwise model0.377
LambdaMART0.381
Medical imaging

Colon Cancer Detection

Flags adenocarcinoma in H&E tissue slides. CLAHE contrast enhancement in OpenCV, then a fine-tuned ResNet.

100% test accuracy on 1,500 held-out images. The case study explains why that number needs a caveat.

Live demo Read case study Code
LLM and retrieval, team project

Symptoma

Predicts likely conditions from symptoms and explains why: random forest, SentenceTransformers, FAISS retrieval and Mistral-7B. Built with Charaneesh at Stevens.

View code

Impact

Numbers from my internship at OYO and my graduate work.

30%Rise in user engagement after launching a mobile-first app
500+Hotels covered by the ranking models I trained at OYO
3.66GPA in the Stevens MS Data Science program

Journey

My education and work experience so far.

Jan 2025 to Dec 2026

MS in Data Science

Stevens Institute of Technology

  • Machine learning, deep learning and applied AI, with projects in RAG and medical AI.
Hoboken, NJ
Jan 2024 to Jul 2024

Technology Engineer Intern

OYO

  • Drove a 30% rise in user engagement by launching a mobile-first application that gave users on-the-go access to key system information.
  • Trained learning-to-rank models with the Ranking and Optimization team to boost high-performing hotels in search results, using features from bookings, clicks, ratings, prices, cancellations and location across 500+ hotels.
  • Evaluated ranking changes with A/B tests and analyzed the results with MySQL queries and dashboards, finding the factors behind higher conversion rates and booking values.
Remote
2020 to 2024

BTech in Computer Science and Engineering

SRM University, AP

  • Specialization in artificial intelligence and machine learning.
Andhra Pradesh, India

Let's build something.

gadusumi@stevens.edu LinkedIn GitHub