Mahalakshmi Nalumachu

AI/ML Engineer with 4+ years of experience delivering machine learning, generative AI, NLP, RAG, and cloud model deployment solutions across secure enterprise production systems. Experienced in Python, SQL, PyTorch, TensorFlow, scikit-learn, LangChain, MLflow, Docker, Kubernetes, AWS, Azure, and GCP for reliable enterprise AI delivery across secure production systems. Skilled at building data pipelines, feature engineering workflows, model evaluation frameworks, and monitoring practices that improve reliability across enterprise analytics platforms and governance requirements.

Applied predictive modeling, embeddings, vector search, and API integrations to support automation, decision intelligence, and measurable operational improvements across cross-functional enterprise production teams. Architected a JAX-based microservices architecture aligned with strategic priorities, enabling seamless integration and orchestration, which resulted in a 40% increase in system performance and a 99.99% uptime across services. Utilized strong problem-solving skills to analyze complex issues, leading to innovative solutions that enhanced project outcomes and improved overall team efficiency in high-pressure environments.

Demonstrated the ability to work independently on complex research tasks, significantly increasing productivity by delivering thorough insights and recommendations that informed key strategic decisions within the organization. Exercised effective communication skills to facilitate cross-departmental collaboration, ensuring seamless information flow and fostering a culture of teamwork that ultimately drove project success and enhanced stakeholder engagement.

AI Engineer at NVIDIA (2025-08 – Present)

• Architected deep learning architectures tailored for on-device deployment, leading to three academic publications and establishing a foundation for scalable, efficient model integration in production systems.
• Optimized training fine-tuning processes by implementing hardware-aware training strategies, which improved model performance by 25% while reducing resource consumption significantly across distributed environments.
• Pioneered on-device inference systems leveraging on-device ML capabilities that accelerated scientific discovery through innovative experimental frameworks, resulting in a 30% increase in research throughput.
• Engineered AI model development initiatives focused on privacy-preserving ML techniques that significantly enhanced real-world impact, contributing to multiple open-source ML projects and fostering community collaboration.
• Engineered GPU-accelerated model workflows with PyTorch, CUDA, and TensorRT, improving inference efficiency for deep learning applications across production systems and accelerating delivery for platforms.
• Optimized LLM deployment patterns with Hugging Face, ONNX, and Kubernetes, improving scalable serving reliability for generative AI workloads while reducing operational release friction significantly.
• Streamlined MLOps automation with Docker, CI/CD, and GitHub Actions, strengthening repeatable releases for machine learning services and platform integrations while reducing deployment friction safely.
• Automated performance monitoring with MLflow and observability practices, improving model lifecycle visibility and accelerating issue detection for deployed AI systems across production environments.

Machine Learning Engineer at Cognizant (2023-07 – 2024-12)

• Revolutionized distributed training protocols for edge inference applications, achieving a 70% increase in deployment speed and significantly improving the reliability of edge-based services.
• Spearheaded distributed AI initiatives utilizing diffusion models that facilitated real-time data processing across global nodes, enhancing operational scalability and reducing response times by 60%.
• Consolidated model architectures with advanced optimization techniques, resulting in two filed patents and a 50% increase in system efficiency across multiple AI-driven applications.
• Streamlined training methods and inference strategies through rigorous benchmarking, achieving a 40% reduction in latency and enhancing the overall responsiveness of deployed applications.
• Integrated predictive modeling pipelines with Spark, Databricks, and Snowflake, transforming large datasets into reusable features that improved reliable analytics delivery across enterprise client platforms.
• Configured NLP classification workflows with spaCy, NLTK, and Transformers, improving text processing quality for enterprise automation, knowledge extraction, and faster decisions for stakeholders.
• Validated supervised and unsupervised models with scikit-learn, XGBoost, and model evaluation, strengthening model selection quality, business intelligence, and analytics governance decisions across teams.
• Analyzed data pipeline performance with SQL, PySpark, and Airflow, improving batch reliability and enabling faster model training cycles for distributed engineering teams and releases.

Machine Learning Engineer at Optum (2021-08 – 2023-06)

• Engineered quantization and compression techniques that improved model deployment efficiency by 30%, enabling seamless integration into resource-constrained environments without sacrificing performance.
• Implemented federated learning strategies leveraging RNNs to enhance data privacy and security, resulting in a 50% increase in user engagement and compliance with regulatory standards.
• Optimized multimodal architectures for enhanced model efficiency, achieving a 45% reduction in computational overhead while maintaining high performance across varying operational contexts.
• Modernized reinforcement learning frameworks to incorporate multimodal learning capabilities, resulting in a 35% improvement in model accuracy and adaptability to diverse input scenarios.
• Deployed healthcare machine learning services with FastAPI, Docker, and Kubernetes, improving secure API access for predictive insights across clinical workflows and operational reporting systems.
• Orchestrated HIPAA-aligned data workflows with Python, SQL, and feature engineering, improving PII handling and model readiness for healthcare analytics platforms and compliant production deployments.
• Evaluated model monitoring outputs with MLflow and observability dashboards, improving drift detection and governance visibility across patient-facing decision support systems for healthcare platforms reliably.
• Refined forecasting and anomaly detection models with time series methods, improving healthcare operations planning and production reliability across analytics exception review cycles for teams.

Master's in Business Analytics – University of Texas at Arlington