AI/ML Engineer for Generative AI Platforms

Manikanteswar Gandrothula

AI/ML Engineer Generative AI & AI Infrastructure

+1-971-***-**** *************.*@*********.*** LinkedIn Open to Relocate SUMMARY

AI/ML Engineer with 4.5+ years of experience building production AI platforms, machine learning systems, and Generative AI applications across finance, healthcare, and AI infrastructure. Expertise in LLM applications, Retrieval-Augmented Generation (RAG), AI platform engineering, distributed training, recommendation systems, NLP, and MLOps, with experience designing production ML pipelines, inference services, and cloud-native AI systems using PyTorch, LangChain, LangGraph, Kubernetes, and AWS. Delivered AI solutions that improved operational efficiency, including automation initiatives generating approximately $30K+ in annual cost savings through optimized ML workflows. SKILLS

Programming & Frameworks: Python, SQL, PySpark, Bash, Scikit-learn, TensorFlow, PyTorch, Hugging Face Transformers, FastAPI Machine Learning & Generative AI: XGBoost, LightGBM, Deep Learning, NLP, Recommendation Systems, Graph Neural Networks (GNNs), Multimodal Learning, Llama, GPT-4, Claude, LangChain, LangGraph, Retrieval-Augmented Generation (RAG), AI Agents, Prompt Engineering, Function Calling, Tool Calling, LLMOps, LLM Evaluation, Llama Guard

Data Engineering & Cloud: Apache Spark, Databricks, Apache Airflow, Kafka, Snowflake, Delta Lake, ETL/ELT, Batch & Streaming Pipelines, AWS (S3, EC2, Lambda, Glue, EMR, SageMaker, CloudWatch), Azure OpenAI MLOps & AI Infrastructure: Docker, Kubernetes, MLflow, CI/CD, REST APIs, Model Serving, Model Monitoring, Feature Stores, Prometheus, Grafana, PyTorch Distributed, DeepSpeed, FSDP, NCCL, CUDA, vLLM, TorchRec Vector Search & Analytics: FAISS, ScaNN, Vector Embeddings, Semantic Search, Hybrid Retrieval, A/B Testing, KPI Analysis, Cohort Analysis, Explainable AI (XAI), Model Governance, HIPAA

PROFESSIONAL EXPERIENCE

Meta Jun 2025 – Present United States

AI/ ML Engineer

Owned the engineering of Meta's internal LLM serving platform, building reusable inference APIs, model versioning, traffic routing, and deployment workflows for Llama 3 models with PyTorch, vLLM, Kubernetes, Docker, and FastAPI, supporting 40+ internal AI applications.

Built a shared RAG platform using LangChain, LangGraph, ScaNN, and hybrid retrieval, enabling engineering teams to search design documents, RFCs, API specifications, and technical knowledge with source-grounded responses, reducing lookup time by 45%.

Designed the distributed training pipeline for Llama fine-tuning with PyTorch FSDP, DeepSpeed, CUDA, mixed-precision training, and Hugging Face Transformers, reducing experiment turnaround from 9 days to under 5 days and improving training reproducibility.

Developed multimodal embedding pipelines that combined text, images, videos, and interaction signals into unified representations for recommendation and content understanding models, improving semantic retrieval quality across large-scale ML workloads.

Established automated LLM evaluation pipelines using Llama Guard, RoBERTa, LangSmith, MLflow, and internal benchmarks to validate factuality, retrieval quality, latency, and policy compliance before production deployments, increasing automated validation coverage to 95%+.

Built high-throughput feature generation pipelines with TorchRec, Spark, Python, and SQL, producing user and content embeddings that powered billions of daily ranking requests and improved feature freshness across recommendation systems.

Improved distributed GPU training efficiency through PyTorch Distributed, NCCL, CUDA, FSDP, and workload scheduling, increasing cluster utilization from 61% to 84% and reducing infrastructure costs for foundation model training. Sentara Health Sep 2024 – May 2025 United States Machines Learning Scientist

Developed deep learning models for 30-day readmission, patient deterioration, and length-of-stay prediction using TensorFlow, ClinicalBERT, structured EHR features, and longitudinal patient records, supporting clinical risk stratification across 2M+ patient encounters.

Implemented AI-assisted clinical documentation workflows using GPT-4, Claude 3.5, and Llama 3 to generate encounter summaries, discharge instructions, and care plan drafts, keeping clinicians in the review loop before patient-facing use.

Developed clinical NLP pipelines with ClinicalBERT, BioBERT, and transformer-based models to extract structured clinical entities from physician notes, increasing structured data availability by 32% for downstream analytics and care management.

Built deployment workflows for ML and GenAI applications using Kubernetes, FastAPI, and MLflow, supporting 15+ production AI services through standardized model releases, scalable inference, and production monitoring.

Delivered HIPAA-compliant AI solutions for clinical documentation, care management, and operational analytics, enabling 400+ clinicians across multiple hospital departments to access patient information more efficiently through AI-assisted workflows. Accenture Jan 2021 – Aug 2023 India

Machine Learning Engineer

Designed probability of default (PD) and credit risk models using Python, XGBoost, LightGBM, Scikit-learn, SQL, and PySpark, supporting automated underwriting decisions across consumer lending portfolios containing 8M+ customer records.

Built a standardized feature engineering framework that consolidated credit bureau data, repayment history, transaction behavior, income verification, and customer attributes into governed feature datasets consumed by multiple risk scoring models.

Automated batch model scoring and validation workflows with Apache Airflow, AWS Glue, Apache Spark, and MLflow, processing 1.5M+ loan applications each month through feature generation, inference, validation, and regulatory reporting.

Productionized machine learning models as containerized REST services using Docker, Kubernetes, AWS SageMaker, FastAPI, EC2, and Lambda, integrating with Loan Origination Systems (LOS) and maintaining inference latency below 250 ms for online credit decisions.

Established MLOps workflows for model versioning, feature drift detection, performance monitoring, and automated retraining using MLflow, Prometheus, Grafana, and AWS CloudWatch, improving operational reliability across production lending systems.

Optimized distributed feature engineering and model training on Apache Spark, reducing end-to-end training time by approximately 7 hours per release cycle and lowering cloud compute costs by an estimated $30K annually through improved resource utilization.

Implemented model governance controls including explainability, validation, audit logging, and version-controlled deployment processes, ensuring compliance with internal risk policies and regulatory requirements throughout the ML lifecycle.

Delivered production ML capabilities for credit underwriting, portfolio risk monitoring, and lending operations, reducing manual credit review effort by approximately 900 analyst hours annually and accelerating credit decision workflows. EDUCATION

Master of Science in Data Analytics Engineering Aug 2023 – May 2025 George Mason University, Fairfax, VA, United States

Contact this candidate