Anmol Dhankhar

• Company Overview: Copilot Insights – Context-Aware Chatbot for Infrastructure Monitoring (ITOM)
• Co-developed the Operations Copilot, an AI-powered assistant leveraging Generative AI for real-time infrastructure monitoring, capable of analyzing system logs and traces from client machines for critical alerts and generate actionable insights.
• Enhanced IT operations efficiency by integrating Natural Language Processing (NLP) for conversational queries, enabling intuitive and faster incident resolution.
• Integrated autonomous AI agents to: Parse logs, identify impacted components, suggest root causes, and recommend remediation steps based on pre-defined SOPs.
• Built dynamic, context-aware dashboards to visualize performance metrics and accelerate decision-making for operations teams.
• Optimized LLMOps workflows for scalable fine-tuning, inference, and retrieval-augmented generation (RAG), ensuring robust multi-agent collaboration.
• Established a human feedback pipeline to enable continuous learning and improve Copilot responses based on real-time user interaction and feedback.
• Copilot Insights – Context-Aware Chatbot for Infrastructure Monitoring (ITOM)

• Led the design, development, and productionization of a Generative AI app for medical writers using RAG.
• Led the implementation of the core backend solution for the Generative AI project.
• Established a framework for streamlined Generative AI project development, reducing completion time.
• Reduced manual effort for medical writers by 70% in drafting non-clinical documents.
• Developed custom chunking logic for PDFs and DOCX, improving LLM content generation.
• Implemented a multi-modal RAG pipeline to handle diverse data types, including tables and charts.
• Integrated OpenSearch Vector DB for faster document retrieval with pre-filter capabilities.
• Received a recommendation on LinkedIn from Pfizer senior leadership for contributions to the Generative AI project.
• Built a predictive model to identify potential healthcare prescribers for Pfizer's drugs in Europe, using tree-based models and logistic regression based on prescription behavior.
• Implemented an MLOps pipeline for automated model training, deployment, and quarterly retraining, with drift monitoring to ensure ongoing model performance.
• Deployed the model on AWS, integrating version control, CI/CD pipelines, and automated validation for scalable and reproducible model deployment.
• Enhanced HCP targeting and adoption ladder criteria using MLOps-driven insights, improving sales team productivity through continuous model updates.

Developed a Prescriber Identification System for Pfizer to boost targeted HCP outreach across multiple drug portfolios

• Led ML solution design and deployment using Dataiku, covering therapeutic areas like Oncology, Cardiology, Pulmonology, etc.
• Built modular training and inference pipelines, enabling full project replication via config changes within weeks.
• Implemented models using Random Forests and Gradient Boosted Trees to classify HCPs based on prescription history and engagement data.
• Addressed class imbalance and used AUC-ROC, F1 score, and Lift@Top-K to evaluate impact.
• Automated retraining every 6 months with live dashboards, prescriber scoring, and feedback loops to track conversion and optimize targeting.
• Enabled ~18% increase in HCP conversion rate and ~30% reduction in outreach effort, significantly improving sales planning and resource allocation.
• Delivered a scalable, reusable ML framework integrated with sales workflows, with per-drug inference endpoints and explainability dashboards.

Developed and deployed a Recommendation System for Dell to automate bulk order replacements for End-of-Life (EOL) hardware products

• Led end-to-end architecture design, model development, and production deployment.
• Built Two-Tower retrieval and LambdaMart re-ranking models to suggest equal or better configurations, reducing manual planner intervention.
• Designed scalable ML pipelines using Amazon SageMaker, Lambda, Step Functions, and Docker.
• Implemented real-time inference using FastAPI, Redis, and Faiss with
• Achieved 30% increase in Recall@10 and 75% reduction in cart replacement time, cutting cart failure rate by 35%.
• Managed a team of 4, coordinated with client SMEs, and iterated based on feedback.
• Utilized contrastive learning, ANN search, and caching strategies to optimize performance and address cold-start issues.

• Collaborated with cross-functional teams to evaluate and optimize system architecture and performance metrics, bridging data engineering and ML operations.
• Architected and implemented a scalable ETL pipeline using AWS Glue, Lambda, and Step Functions, streamlining large-scale data ingestion and processing.
• Enhanced machine learning workflows by refining data preprocessing and feature engineering techniques to improve input dataset quality.
• Optimized model performance through systematic hyperparameter tuning and algorithmic improvements, ensuring efficiency and scalability.
• Deployed and monitored robust machine learning solutions on AWS SageMaker, maintaining high reliability and performance in production.
• Established automated monitoring systems to detect and mitigate data, model, and concept drift, ensuring continuous model health.
• Developed and maintained CI/CD pipelines for seamless model versioning, deployment, and regular retraining cycles.
• Stayed at the forefront of AI/ML advancements, consistently recommending and implementing innovative technologies to optimize the machine learning lifecycle.

• Solved a classification problem using tree-based models to predict users eligible for targeted advertisements.
• Automated production-level model monitoring to detect data, model, and concept drift.
• Managed model versioning and configuration, ensuring robust and reliable deployments.
• Executed model refreshes through A/B testing and shadow testing to validate performance improvements.
• Utilized Docker and Kubernetes for automated deployment, scaling, and management of containerized applications across clusters.