➼ AI Predictions for Key Refinery Metrics:Led and managed a team in delivering AI prediction models tailored to enhance operations and predict vital metrics at specific refinery locations.➼ AI Models for Fuel Quality and Composition:Managed teams in the development and deployment of AI models predicting fuel quality, density, and key content components, ensuring operational efficiency at a major refinery.➼ Agile Methodology & Certifications:Certified with PSM 1 (Professional Scrum Master) and adeptly implement Agile methodologies in project management.Regular participant in daily stand-ups, sprint reviews, sprint plannings, and backlog grooming, driving the success of projects through timely and efficient deliveries.➼ Presentation at an World Economic Forum:Showcased our innovative AI models at a World Economic forum Light house awards, emphasizing our commitment to pioneering industry solutions.➼ Unified Monitoring Platform for Gas Plants:Championed the development of a unified AI-driven platform to streamline on boarding and monitoring for multiple gas plants, enhancing oversight of chemical interactions in various equipment.➼ AI Predictions for Gas Plant Content:Successfully spearheaded a team in creating an AI model predicting specific chemical content, enhancing the operations of a major gas plant.➼ AI-Driven Executive Reports:Orchestrated a specialized team leveraging state-of-the-art LLMs and deep learning techniques to produce a daily executive report, encapsulating crucial insights and data for company leadership.

➼Designed, developed, and maintained a Propensity Model for credit card lead recommendations, optimizing for imbalanced data using techniques such as logistic regression, boosting, and bagging algorithms.➼Led the migration of data from Oracle DB to the Hadoop ecosystem.

➼ Spearheaded rapid prototyping for AI-powered summarization of legal documents, dramatically reducing lawyers' reading time by pinpointing new rules and deploying a keyword extraction algorithm to spotlight key phrases.➼ Took the lead on document tagging to determine the relevance of rules in documents to associated forms. Established a semantic word-matching dictionary between documents and forms, delivering substantial savings in lawyer time and company costs.➼ Enhanced, developed, and recalibrated the production code for the LegalBill View Analyser, serving four clients, to ensure correct billing by law firms based on attorney workload.➼ Pioneered an AI solution to generate rationale for adjustments recommended by the LegalBill View Analyser (LBA) model. This innovation is now live and under evaluation by LBA clients for feedback.➼ Actively revamping the existing LBA machine learning framework, harnessing Deep Learning and predefined rules to heighten precision and recall. With the integration of simple Dense layers and Doc2Vec in lieu of Word2Vec, we've observed a 3.5% surge in both precision and recall.Technologies Utilized: Json, Oracle Db, Reg Ex, TF-IDF, Text Rank, K-means Clustering, word embedding, word2vec, doc2vec, Python, distance algorithms, LDA (Topic modelling), Random Forest, XGBOOST, Deep learning (Dense layers, CNNs, Bi-Directional LSTMs).

➼ Developed a highly efficient RPA system using Python, tailored for low-resource devices such as Raspberry Pi, to automate the scraping, cleaning, and processing of data for over 1800 stocks—covering daily highs/lows, advances/declines, bulk deals, quarterly tax statements, balance sheets, and cash flows.➼ Partnered with the team to engineer advanced HFT algorithms for real-time buy/sell triggers on live data for 1800+ stocks. Utilized technical tools like SMI, MDA, oscillators, Bollinger Bands, Pivot Points, RSI, and Super Trend, achieving decision times as swift as ~3ms for all stocks.➼ Pioneered a stock news sentiment classifier and a stock recommendation system by harnessing cutting-edge deep learning methodologies.➼ Enhanced the efficiency of HFT algorithms by integrating the outputs of the news sentiment and recommendation engine, yielding buy/sell triggers that outperformed traditional models by 25%.Technologies Utilized: Python, Selenium, Beautiful Soup (BS4), lxml, pandas, numpy, Keras, Doc2Vec, Gensim.

➼ Spearheaded rapid prototyping and showcased applications geared towards real-world evidence utilization within the pharmaceutical sector.➼ Designed and presented innovative models to harness social media data, capturing patients' authentic treatment experiences — encompassing drug efficacy, safety, and overall impact. Achieved a 25% improvement in scoring time and bolstered accuracy by 5% by synergizing rule-based systems with machine learning techniques.➼ Successfully orchestrated the migration of both new and incremental data sets from Teradata to HDFS employing Pyspark.Technologies Leveraged: HDFS, Sqoop, Pyspark, Hive, Impala, Python, Shell Scripting, Machine Learning.