Ankit Srivastava

Lead DS team, providing mentorship, & overseeing project lifecycles from problem definition to deployment. Lead weekly team meetings to align objectives, discuss ongoing projects, & strategize improvements. Scraped, cleaned, analyzed, & fine-tuned LLMs utilizing PEFT (LoRA) for various applications of our product including Sentiment Analysis, Aspect Analysis, Spam Detection, Topic Modelling, & Text Summarizing. Developed natural language query to SQL using langchain function calling & utilized vector DB to search context. Utilized RAG map-reduce technique for summarizing topics and providing an overall summary for queried texts.

Led DS team & spearheaded ML-driven solutions to automate manual processes, and enhance software efficiency. Developed model for the Catalogue Automation (182 categories), utilizing XGBClassifier (fit on tabular data with sentence embeddings & Word2Vec for description) to categorize products with 92% accuracy, and 82.3% F-Score. Analyzed data for Item Return Shipment, employing predictive modeling techniques to anticipate returned items.

Collaborated with the NDR news channel team to conduct multi-class (9 classes) classification on their news website, achieving accuracy 93% & F-Score 82% using a Neural Network-based approach. Led the development of a multimodal application for second language learners, integrating complex word identification models by fine-tuning RoBERTa, resulting in a model accuracy of 95% and an F-Score of 87%. Played a key role in data collection, feature engineering, and model development, demonstrating expertise in data science techniques and advanced neural network architectures.

Developed and maintained Hospitality Property Management System from requirements analysis to development. Analyzed data & developed a machine learning model for Customer Market Segmentation using clustering algorithms (K-means, DBSCAN, HAC), achieving a Silhouette Score of 0.57.

Managed client interactions (from USA/UK) and conducted comprehensive requirement analysis. Worked as full-stack developer using Progress 4GL, delivering new features & optimizing legacy code. Successfully resolved a decade-old issue stemming from legacy code, showcasing problem-solving skills. Mentored junior developers and provided onboarding training for experienced team members.

Developed an E-Commerce web application prototype for client utilizing Java and Oracle Database.