siddharth srivastav

I’m a Data Scientist focused on turning complex, messy data into reliable, production-ready AI/ML and Generative AI workflows. I specialize in building scalable data pipelines, optimizing preprocessing systems, improving model performance, and deploying intelligent AI solutions that deliver consistent, interpretable outputs for real-world applications.

In my most recent role, I built and improved preprocessing workflows for high-volume datasets and supported algorithm-integrated systems for automated classification, extraction, and report generation. I collaborated cross-functionally to productionize AI-driven workflows by integrating ML models with scalable backend services, improving automation efficiency and reducing manual review efforts.

I have hands-on experience with Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and Agentic AI systems, including building intelligent workflows that combine LLM reasoning, vector databases, semantic search, prompt engineering, and automated decision-making agents. I have worked on optimizing domain-specific LLM applications, improving retrieval accuracy, enhancing context generation, and designing AI pipelines capable of handling complex user queries and multi-step tasks.

Previously, I worked on predictive healthcare modeling, improving patient risk analytics by ~20% using ensemble learning techniques such as Random Forest and Logistic Regression. I performed exploratory data analysis on 10M+ records using Python (Pandas, NumPy) and SQL, developed interactive dashboards using Python and Power BI across multiple projects, and achieved AUC scores of 0.85+ through model optimization, feature engineering, and evaluation of supervised and unsupervised learning models.

My experience also includes fine-tuning domain-specific LLMs on large-scale datasets, optimizing NLP pipelines, and reducing model training time through efficient preprocessing and scalable ML workflows. I enjoy research-driven problem solving—whether it involves developing Agentic AI assistants, implementing RAG-based knowledge systems, experimenting with reinforcement learning models, or creating automated feature engineering pipelines using Scikit-learn and modern ML frameworks.