Abhishek Yadav

Implemented a Logistic Regression classification model using past campaign data to optimize luxury real estate conversion through recall prioritization. Employed OpenAI GPT-3.5 API for few-shot listing description generation. Experimented with OpenAI GPT-3.5, OpenAI Embeddings, LangChain, and LanceDB/FAISS vector Database to prototype a RAG-based AI chatbot.

Developed a highly accurate CalibratedClassifier ML model with Light Gradient Boosting as base estimator to enable retail agents in India to prioritize leads and boost the important metric of I2P from 14% to an impressive 18%. Implemented Price fraud detection utilizing XGBRegressor and employed fraud detection using Text classification with LDA(Latent Dirichlet Allocation) topic modeling for the two-wheelers category within Trust and Safety.

Developed a scalable feature engineering pipeline that generates 60+ technical indicators for any trading asset following the OHLCV pattern. Used Adaboost and RandomForestClassifier models from scikit-learn with the triple barrier method to predict price movements and achieved positive Sharpe ratio for strategies by backtesting models using TimeSeriesSplit.

Developed an automated Django web application for AMG IAP, streamlining operations by 95% through process automation and report generation. Created an analytics platform utilizing streamlit, plotly, seaborn, and wordcloud. Mobile app; expense tracking, investment advice, and query management.

Created a deduplication model via fuzzy-wuzzy matching and k-nearest neighbors, alongside a sentiment analysis model. Utilized BeautifulSoup, lxml, tika, and Selenium within NiFi for diverse pharma domain data mining. Awarded Certificate of Excellence for outstanding performance as Intern.