Mounisha Vanga

<p>Designed and deployed enterprise LLM applications that automated knowledge retrieval and reduced manual research effort by 40% across medical and business teams.</p><p> Built Retrieval-Augmented Generation (RAG) pipelines using embedding models and vector databases including FAISS, Pinecone, and ChromaDB, reducing irrelevant AI-generated outputs by 35%.</p><p> Built agentic AI workflows in LangChain and LangGraph with tool calling, structured outputs, retries, and failure handling for multi-step enterprise tasks.</p><p> Built and deployed Model Context Protocol (MCP) servers exposing internal knowledge bases, vector stores, and enterprise APIs as standardized tools for LLM applications.</p><p> Implemented MCP clients and tool-calling workflows querying document repositories, SQL databases, and internal REST services with scoped access controls and audit logging.</p><p> Integrated AI APIs including Azure OpenAI and Anthropic Claude into enterprise applications through FastAPI backend services and REST APIs.</p><p> Engineered prompts, system instructions, and structured outputs that improved response consistency and reduced hallucination rates.</p><p> Fine-tuned open-source LLMs on domain-specific datasets to improve task performance and domain terminology understanding.</p><p> Developed document ingestion and semantic search pipelines for clinical reports, research papers, PDFs, and regulatory documents.</p><p> Developed NLP workflows for classification, entity extraction, summarization, and semantic search across 100K+ healthcare and research documents.</p><p> Deployed containerized AI services on Kubernetes with autoscaling, resource management, and controlled rollouts.</p><p> Built CI/CD pipelines automating build, test, evaluation gates, deployment, and rollback for AI applications and backend services.</p><p> Provisioned cloud infrastructure through Terraform and infrastructure-as-code, improving consistency across environments.</p><p> Implemented monitoring and observability across AI services covering logging, latency, errors, token usage, drift, and alerting.</p><p> Built automated evaluation frameworks measuring accuracy, relevance, latency, and factual consistency before each production release.</p><p> Performed static and dynamic code analysis using SonarQube, Pylint, and Bandit to detect vulnerabilities, code smells, and dependency risks before release.</p><p> Implemented AI governance including monitoring, access controls, security policies, and responsible AI validation.</p><p> Used Claude Code, GitHub Copilot, and Cursor for AI-assisted development, refactoring, automated code review, unit test scaffolding, and documentation.</p><p> Optimized inference pipelines through batching, caching, and quantization, reducing LLM processing cost by 25% while maintaining performance.</p><p> Supported production AI applications through incident troubleshooting, root-cause analysis, and reliability improvements.</p>

<p>Developed machine learning models for fraud detection, risk scoring, and customer behavior prediction, improving detection accuracy by 20%.</p><p> Designed and deployed production models and services on AWS SageMaker and Azure Machine Learning with managed endpoints and IAM-based access.</p><p> Developed REST APIs and containerized microservices delivering real-time predictions into enterprise banking and insurance applications.</p><p> Built automated CI/CD pipelines in Jenkins and GitLab CI to containerize, test, and deploy applications to Docker and Kubernetes with zero-downtime releases.</p><p> Managed infrastructure as code using Terraform and Helm to provision reproducible training and inference environments across development, staging, and production.</p><p> Established blue-green and canary deployment strategies with automated rollback and Prometheus/Grafana observability, reducing incident recovery time.</p><p> Embedded DevSecOps controls into delivery pipelines, integrating SAST, DAST, secret scanning, dependency checks, and container image scans using SonarQube, Snyk, and Trivy.</p><p> Automated secrets management and least-privilege access using HashiCorp Vault and IAM role-based policies, maintaining compliance with PCI-DSS and internal risk standards.</p><p> Built automated feature engineering pipelines processing millions of financial transactions to support real-time predictions.</p><p> Engineered streaming and batch data pipelines with Spark and Kafka feeding production services.</p><p> Automated model training and deployment workflows, reducing manual release cycles by 50%.</p><p> Implemented production monitoring tracking model drift, data quality, latency, and prediction performance with alerting.</p><p> Conducted A/B testing and statistical evaluation to measure solution effectiveness before full production rollout.</p><p> Troubleshot production incidents through root-cause analysis and resource tuning, improving reliability and latency.</p><p> Documented workflows, architecture, and validation results supporting governance and audit requirements.</p>

<p>Developed production backend services and batch processing components in Java, integrating machine learning outputs into enterprise risk, settlement, and reporting platforms.</p><p> Implemented TypeScript-based API clients and shared type definitions that standardized integration between data science services and enterprise web applications.</p><p> Built interactive analytics and model monitoring interfaces using JavaScript and TypeScript, enabling business users to explore risk metrics and model performance.</p><p> Led end-to-end machine learning projects from data exploration and feature engineering through deployment and business adoption.</p><p> Developed predictive models for risk analysis, anomaly detection, and operational forecasting, improving decision accuracy by 25%.</p><p> Built classification, regression, and time-series forecasting models using large-scale financial datasets.</p><p> Designed feature engineering strategies that improved model performance and reduced false positive rates.</p><p> Analyzed millions of financial records using Python, SQL, and Spark to uncover actionable insights.</p><p> Built distributed processing pipelines on Spark and Databricks with orchestration managed through Airflow.</p><p> Improved data quality through validation rules, cleansing workflows, and automated checks.</p><p> Collaborated with engineering teams to productionize machine learning models into enterprise systems.</p><p> Created automated dashboards and reporting solutions using Power BI and Tableau for executive decision-making.</p><p> Maintained model documentation, validation reports, and compliance artifacts for regulated environments.</p><p> Mentored junior data scientists on modeling approaches, data analysis techniques, and best practices.</p><p> Presented analytical findings and recommendations to senior leadership teams.</p>

<p>Developed predictive models for loan approval, customer retention, and borrower risk analysis using historical lending data.</p><p> Built classification and regression models that improved customer targeting accuracy by 20%.</p><p> Performed exploratory data analysis on customer and financial datasets to identify business improvement opportunities.</p><p> Designed customer segmentation models supporting personalized marketing campaigns.</p><p> Applied feature engineering techniques to improve model accuracy and predictive performance.</p><p> Automated recurring business reports and data pipelines using SQL and Python, reducing manual reporting effort by 30%.</p><p> Developed C# and .NET Core services and ASP.NET Core Web APIs that exposed model scores to loan origination applications.</p><p> Integrated Python-trained models into .NET application workflows through REST endpoints and scheduled scoring services.</p><p> Collaborated with .NET application teams on data contracts, testing, and deployment of model-driven features into production.</p><p> Standardized reproducible training code with version control, environment management, and documented procedures.</p><p> Evaluated model performance using accuracy, precision, recall, and ROC analysis.</p><p> Created dashboards and analytical reports using Tableau and Power BI for business stakeholders.</p><p> Collaborated with business teams to translate analytical findings into actionable recommendations.</p><p> Documented analytical methodologies, assumptions, and model outcomes for future enhancements.</p>

<p>Analyzed manufacturing and quality datasets to identify operational trends and improve production visibility.</p><p> Developed SQL queries for reporting, operational analysis, and business performance tracking.</p><p> Built Power BI and Tableau dashboards monitoring production KPIs, quality metrics, and inventory performance.</p><p> Automated recurring Excel reports using VBA, reducing manual reporting effort by 40%.</p><p> Performed statistical analysis to support quality investigations and manufacturing process improvements.</p><p> Improved data accuracy by validating and reconciling information across ERP and manufacturing systems.</p><p> Participated in the full SDLC for reporting and analytics applications, contributing to requirements, design, development, testing, deployment, and post-release support.</p><p> Authored functional and technical specification documents, source-to-target mappings, and test cases as formal SDLC deliverables for validated pharmaceutical systems.</p><p> Executed unit, integration, and user acceptance testing (UAT) for BI reports and data workflows, coordinating defect resolution with development teams.</p><p> Worked within Agile and Waterfall methodologies, participating in sprint planning, stand-ups, backlog grooming, and release readiness reviews.</p><p> Supported change control, version management, and release documentation in line with GxP and internal quality standards.</p>