AI Projects

Regime selection

  • Predicted regimes in S&P index price using Hidden Markov Gaussian mixture models.

Kinship Verification

  • Detected Kinship with 85.5% accuracy on Families in Wild dataset between a pair of images using Siamese Networks , with features getting transfer learned from VGG16 network with adam optimizer and binary cross entropy loss function.

Aspect Based Sentiment Analysis(under Prof. Aruna Malapati)

  • Split sentences into multi-level clusters based on their multiple subjects in sentence using Optics Clustering
  • Extracted aspects and associated words relevant in the sentence using attention based
    modeling of sentence
  • Determined Sentiment polarity of the entire sentence both and also of concerned aspect term by using Long Short term memory network


  • Created recommender systems for movies using the MovieLens dataset by implementing user-user and item-item collaborative filtering algorithms.
  • Also explored SVD and CUR decomposition which reduced the prediction time by 400%
Recommender System TechniqueRoot Mean Square Error (RMSE)Precision on top KSpearman Rank CorrelationTime taken for prediction (secs)
Collaborative2.033519 (item) 2.1502(user)0.6016 (item) 0.584474(user)0.99999975(item) 0.99999972 (user)211.979 (item) 272.817 (user)
Collaborative along with Baseline approach0.939036 (item) 1.005434 (user)0.62865586 (item) 0.64406025 (user)0.999999947 (item) 0.99999939 (user)313.3369 (item) 273.2009(user)
SVD with 90% retained energy1.030.65280.999999999839361.49

Search Engine for Research Papers

  • A vector-space tf-idf model based research paper search and discovery engine using Arxiv data
  • Used Trie data structure to empower autofill for the research paper search engine

Indian Election Prediction using Twitter

  • Predicted the correct results of india’s lok sabha election by doing attention based sentiment analysis of tweets
  • Created a scoring mechanism associated with different types of twitter activities with retweet getting less weightage compared to original tweets and people with more followers getting more weightage compared to others.

Corona Vitals

  • Generated base vital signs data based on the “Novel coronavirus infection during the 2019–2020 epidemic” in Sichuan Province, China. Analyzed vitals signs data using deep learning in keras for suspected COVID patients
  • Presented at BARDA Industry Day Lightning Talks by US Health Department

Shopee Price Match Guarantee

  • Worked on detecting if two products are similar on the kaggle competition using Siamese Network with VGG 16 encoding.

Haar Cascade

  • Trained a cascade to recognize human faces and eyes using OpenCV


  • Enhanced educational videos with natural language processing and computer vision by providing text summarization, POS tagging & NER for general knowledge and completed the fully functional prototype in form of a website within just 36 hours at Hack Harvard


  • Anterior Chamber Depth and Corneal Thickness at Srujana Innovation Center
  • Project: Constructed a miniature prototype that helps in early detection of Glaucoma Disease by using the Slit width image of an eye
  • Worked on Image segmentation and Image analysis techniques using Python, OpenCV, Raspi3, matplotlib to calculate Anterior Chamber Depth and Corneal thickness.
  • Awarded the best team award in 5th Engineering the eye Hackathon organized by MIT MEDIA LAB and Srujana Innovation

Market Basket Analysis

Market Basket Analysis and Mining Association Rules
Project: Proposed a model that analyzes customer behavior iterating
through different range of values for support and confidence
• Used Apriori Principle and FP Tree for candidate pruning and rule
• Formulated a model that gives insight about customers shopping
patterns and also suggest the viable offers.
• Tools: C++, FpTrees, Apriori Principle, Association Rule Mining

Detect Human Trafficking

  • Tested out to Abnormal event detection is videos using generative adversarial nets.
  • Won the Tryst IIT-Delhi , 2019 competition.

Portable eye profiler

  • creating a portable eye profiler using a modified VR setup. The goal was to prototype and demonstrate the possibility of using simple paper cutouts for advance medical screening.
  • Won the $1500 cash prize at Zeiss Hacks in Bangalore


  • Implemented a algorithm that recognize the object and follows it with the help of servo blasters and raspberry pi

Parkinson Disease detection

  • Detected Parkinson disease by analyzing the spiral handwriting of patients.
  • Explored several classification algorithms such as Logistic Regression, Random Forest, SVM etc. The best results were obtained using SVM.
Accuracy100 %

Optimal Path in Hyderabad

  • Applied A star algorithm to find optimal path between two given places in hyderabad.

Predicting Google Stock Price using LSTM

  • Predicted stock prices from historical data using LSTM


ML in C++, Java


Concussion Detection – Medhacks

  • Detected concussion using an mobile application in OpenCV

Background Subtraction

  • Able to disintegrate foreground from background to capture moving objects in opencv python

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: