Home

Vijay Manikanta - AI/ML Engineer

[email protected]

Location: Denver, Colorado, USA

Relocation: Yes

Visa: Green card

Resume file: ML resume_1779457549682.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

Vijay Manikanta Paturi | ML/AI Engineer | Data Scientist | Data Engineer [email protected] | +1 213-802-9355 www.linkedin.com/in/v-manikanta-p/ PROFESSIONAL SUMMARY Strategic AI Visionary & Lead Engineer: A career defined by 10+ years of technical evolution, transitioning from foundational frequentist statistics and predictive modeling architecting cutting-edge Agentic Workflows, Multi-Modal Generative AI, and Self-Healing MLOps ecosystems. Generative AI & LLM Orchestration: Expert in the end-to-end lifecycle of Large Language Models, including Instruction Fine-Tuning using PEFT (LoRA/QLoRA), Prompt Engineering (Chain-of-Thought, ReAct), and the development of sophisticated Retrieval-Augmented Generation (RAG) pipelines. Proficient in optimizing inference for private, enterprise-grade LLM deployments using vLLM, Triton Inference Server, and Model Quantization (GGUF, AWQ). Production-Grade "AI-as-a-Service": Highly specialized in the "Productionization" of experimental research. Expert at bridging the gap between Jupyter-based prototyping and robust, microservices-oriented software engineering. Proven ability to deploy sub-100ms latency inference engines within high-compliance sectors including Healthcare (HIPAA), Finance (SEC), and Automotive safety standards. Advanced MLOps & Infrastructure: Distinguished track record in architecting MLOps 2.0 frameworks, integrating Feature Stores (Feast, SageMaker Feature Store), Automated Model Retraining Loops, and Model Observability for drift and bias detection. Expert in container orchestration using Docker and Kubernetes (K8s/EKS/GKE), managing distributed GPU/TPU workloads to process 100TB+ daily data streams. Big Data & Data Engineering Excellence: Master of the "Data-Centric AI" philosophy, building scalable ETL/ELT pipelines using PySpark, Apache Spark Streaming, and Kafka. Deep expertise in modern data lakehouse architectures (Delta Lake, Snowflake, BigQuery) and vector databases (Milvus, Pinecone, Weaviate) to support high-dimensional semantic search. Computer Vision & NLP Deep Dive: Extensive experience in developing custom Transformer-based architectures, CNNs (YOLOv11, ResNet), and RNNs for real-time object detection, medical imaging diagnostics, and multi-lingual sentiment analysis. Expert in leveraging Hugging Face, PyTorch, and TensorFlow to build proprietary models that outperform baseline open-source implementations. Cross-Functional Leadership & AI Governance: Adept at translating complex Explainable AI (XAI) metrics into actionable ROI for C-suite stakeholders. Led multi-disciplinary teams through the SDLC for AI, ensuring adherence to AI Ethics, security guardrails, and rigorous data governance protocols. Quantifiable Business Impact: Consistently delivering high-impact solutions that have reduced operational overhead by up to 60%, improved model accuracy through Ensemble methods and Hyperparameter Tuning (Optuna) by 70% and accelerated the AI-to-Market lifecycle by 40% via CI/CD automation. Technical Versatility: A "Polyglot" engineer proficient in Python (Expert), R, Java, and SQL, with a deep understanding of cloud-native services across AWS, Azure, and GCP, including SageMaker Pipelines, Vertex AI, and Azure ML Studio. Core Professional Mission: Committed to solving the world s most complex industrial challenges by integrating Deep Learning, Agentic Intelligence, and Massive Data Scale into transparent, ethical, and highly performant digital solutions. TECHNICAL SKILLS & AREA OF EXPERTISE Generative AI & LLMs Architecture: RAG (Retrieval-Augmented Generation), Agentic Workflows, Multi-modal LLMs, Chain-of-Thought (CoT), ReAct. Frameworks: LangChain, LlamaIndex, Haystack, AutoGen, CrewAI, LangGraph. Models: Llama 3.1/3.2, GPT-4o, Claude 3.5 Sonnet, Mistral, BERT, T5, CLIP, Whisper. Optimization: PEFT, LoRA, QLoRA, Quantization (GGUF, AWQ, EXL2), Model Pruning, vLLM, SGLang. Vector Search & Embeddings Vector DBs: Pinecone, Milvus, Weaviate, FAISS, ChromaDB, Qdrant, Azure AI Search. Retrieval: Hybrid Search (BM25 + Vector), Semantic Re-ranking, Cross-Encoders, Cohere Rerank. Machine Learning & Deep Learning Deep Learning Architectures: Graph Neural Networks (GNN), Self-Attention Mechanisms, Multi-Head Attention, CNNs (YOLO, ResNet), Autoencoders, GANs. Frameworks: PyTorch, TensorFlow, Keras, JAX, Caffe2. Algorithms: Gradient Boosting (XGBoost, CatBoost, LightGBM), Random Forest, SVM, KNN, H2O.ai. Advanced ML: Reinforcement Learning (PPO, DQN), Bayesian Optimization, Optuna, Ray Tune. Computer Vision (CV) Architectures: Transformers (ViT), CNNs (YOLOv8-v11, ResNet, EfficientNet, MobileNet). Processing: OpenCV, MediaPipe, Scikit-Image, PIL, Image Augmentation (Albumentations). Tasks: Object Detection, Instance Segmentation, Facial Recognition, OCR (Tesseract, PaddleOCR). Natural Language Processing (NLP) Architectures: Attention Networks, Transformers (Encoder/Decoder), RNNs (LSTM, GRU), Bidirectional Encoder Representations (BERT), RoBERTa. Libraries: Hugging Face Transformers, spaCy, NLTK, Gensim, Stanza. Tasks: Named Entity Recognition (NER), Sentiment Analysis, Text Summarization, Topic Modeling (LDA). MLOps & AI Infrastructure Platforms: AWS SageMaker (Pipelines, Feature Store), Google Vertex AI, Azure ML Studio, Databricks. Orchestration: Kubernetes (K8s/EKS/GKE), Docker, Helm, Kubeflow, Apache Airflow, Prefect. Tracking & Versioning: MLflow, Weights & Biases (W&B), DVC (Data Version Control), ClearML. Monitoring: Prometheus, Grafana, Evidently AI (Drift Detection), Fiddler (Explainable AI). Data Engineering & Big Data Distributed Computing: PySpark, Apache Spark, Spark Streaming, Hadoop, Hive, MapReduce. Streaming: Apache Kafka, AWS Kinesis, RabbitMQ, Confluent. Warehousing: Snowflake, Google BigQuery, Amazon Redshift, Delta Lake, Databricks Lakehouse. Cloud Computing & DevOps AWS: Lamb da, Glue, S3, EC2, IAM, Bedrock, Kendra, Step Functions. GCP: BigTable, Cloud Functions, Pub/Sub, Dataflow. IaC & CI/CD: Terraform, CloudFormation, GitHub Actions, Jenkins, CircleCI. Programming & Backend Languages: Python (Expert), SQL (T-SQL, PL/SQL), R, Java, Scala, C++, Bash. Backend: FastAPI, Flask, Django, Pydantic, Celery, Redis, RabbitMQ. Databases & Storage Relational: PostgreSQL, MySQL, MS SQL Server, Oracle. NoSQL: MongoDB, Cassandra, DynamoDB, Elasticsearch, Neo4j (Graph). Analytics & Visualization BI Tools: Tableau, Power BI, Looker, Grafana. Libraries: Matplotlib, Seaborn, Plotly, Streamlit, Gradio (for AI Demos). Governance & Security AI Ethics: Bias mitigation, Fairlearn, AI Explainability (SHAP, LIME), NeMo Guardrails, HIPAA/GDPR Compliance. PROFESSIONAL EXPERIENCE Senior AI Engineer | Goldman Sachs, New York, NY | March 2024 Present Architecting and deploying autonomous Agentic AI workflows using LangGraph, CrewAI, and Llama 3.1, enabling multi-step reasoning and automated decision-making for complex regulatory and financial audit cycles. Leading the development of enterprise-grade Retrieval-Augmented Generation (RAG) systems, integrating Hybrid Search (Vector + Keyword), Cross-Encoders, and Semantic Re-ranking to achieve 95%+ retrieval accuracy across millions of unstructured documents. Designing and managing private, secure LLM infrastructure on AWS EKS using vLLM and NVIDIA Triton Inference Server, implementing Model Quantization (AWQ/GGUF) to optimize GPU memory and increase request throughput by 3x. Orchestrating the "Productionization" of experimental AI research by building robust, microservices-oriented software wrappers around PyTorch and TensorFlow models, ensuring sub-100ms latency for real-time inference. Fine-tuning Large Language Models (LLMs) including Llama 3, Mistral, and Claude using Parameter-Efficient Fine-Tuning (PEFT) techniques such as LoRA and QLoRA for domain-specific medical and automotive corpora. Implementing AI Governance protocols and safety guardrails using NeMo Guardrails to prevent hallucinations and ensure PII masking within customer-facing generative interfaces. Optimizing model inference for real-time applications by implementing SGLang for continuous batching, reducing time-to-first token (TTFT) by 150ms for agentic interfaces. Developing specialized AI agents for "Reasoning," "Tool-use," and "Final Verification," effectively reduce human audit cycles from 15 days to 4 hours. Translating complex algorithmic outcomes into strategic business ROI for C-suite stakeholders, aligning technical AI roadmaps with corporate growth and operational efficiency KPIs. Senior ML / AI Engineer | CVS Health, Irving, TX | September 2020 January 2024 Established an enterprise-wide MLOps 2.0 framework using AWS SageMaker, MLflow, and Kubeflow, automating the lifecycle of 300+ models from experimentation to continuous deployment and monitoring. Developed custom Computer Vision solutions using YOLOv11 and Vision Transformers (ViT) for high-speed industrial applications, achieving 98.4% precision in detecting anomalies across high-resolution video streams. Engineered multi-modal diagnostic engines for healthcare applications, fusing CNN (ResNet) imaging features with RNN (LSTM/GRU) based patient timeline data to provide comprehensive predictive insights. Managed massive 100TB+ data streams through PySpark and Delta Lake architecture, implementing Feature Stores (SageMaker/Feast) to provide low-latency feature retrieval for real-time model serving. Implemented Explainable AI (XAI) frameworks using SHAP and LIME to provide interpretability for black-box models, ensuring compliance with global AI ethics and transparency standards. Developed self-healing pipeline monitoring using Evidently AI, Prometheus, and Grafana, triggering automated retraining loops when significant feature or concept drift is detected in production. Optimized deep learning models for edge and cloud deployment through TensorRT and ONNX conversion, reducing inference costs by 40% while maintaining model performance thresholds. Led cross-functional teams of data scientists and DevOps engineers to implement CI/CD for Machine Learning, ensuring rigorous testing, versioning of data with DVC, and seamless model rollbacks. Led the migration of legacy RNN/LSTM sentiment analysis models to Transformer-based architectures (BERT, RoBERTa), improving multi-lingual text classification F1-scores by 25%. Conducted kernel profiling and memory tuning for GPU optimization using CUDA and Habana SynapseAI SDK, achieving significant throughput gains for distributed training workloads. Partnered with product and design teams to embed AI models into mobile app prototypes, utilizing Prophet and ARIMA for time-series forecasting of user behavior and patient deterioration. Machine Learning Engineer / Data Engineer | Vistan NextGen Hyderabad | January 2017 July 2020 Engineered and maintained robust ETL pipelines using Python and PySpark to ingest and process high-velocity telemetry data from over 10,000 IoT devices across smart industrial environments. Optimized data ingestion workflows to handle 50GB+ of daily streaming data, utilizing Apache Kafka for real-time message queuing and Apache Airflow for complex DAG orchestration. Designed and implemented scalable data architectures by integrating AWS IoT Core, Apache Hive, and MongoDB, enabling real-time monitoring and predictive fault detection for mission-critical hardware. Developed custom deep learning pipelines using CNNs for automated visual inspection on production lines, significantly reducing manual QA overhead and improving defect detection rates. Built and deployed secure, high-performance REST-based connectors and FastAPI wrappers to serve model predictions to internal operational dashboards and mobile prototypes. Collaborated with senior data scientists to build a centralized Feature Store, ensuring feature consistency between training and serving environments and reducing model development lead times by 30%. Integrated multi-source datasets from PostgreSQL, S3, and NoSQL databases into a unified Data Lake architecture, facilitating large-scale exploratory data analysis and hypothesis testing. Automated recurring data validation and quality check protocols using SQL and Python scripts, identifying schema inconsistencies and reducing data-related production issues by 25%. Designed and maintained interactive visualization dashboards in Power BI and Looker to monitor device anomalies, optimize energy consumption, and assist in capacity planning. Supported the operationalization of early-stage NLP models for log file analysis, utilizing RNNs (LSTMs) to predict system failures based on sequential error patterns. Conducted rigorous performance profiling of data jobs, utilizing Spark UI and CloudWatch to identify bottlenecks and reduce cloud computing costs by 15% through query optimization and resource right-sizing. Partnered with DevOps teams to containerize ML workloads using Docker, establishing the foundation for consistent model behavior across development, staging, and production environments. Documented entire data lineage and transformation logic, providing a clear audit trail for compliance and simplifying the onboarding process for new engineering hires. Junior Data Scientist | Inovalon Hyderabad | January 2015 December 2016 Spearheaded comprehensive exploratory data analysis (EDA) using Python (Pandas, NumPy) and R to identify seasonal demand patterns and anomalies in large-scale retail and financial datasets. Developed and deployed foundational predictive models, including Linear Regression, Logistic Regression, and Decision Trees, to forecast customer churn and lifetime value, achieving a 15% increase in retention accuracy. Performed rigorous statistical analysis, including Hypothesis Testing (A-B Testing), ANOVA, and Chi-square tests, to validate business assumptions and optimize marketing campaign performance. Architected and optimized complex SQL queries (T-SQL/PL-SQL) to extract, transform, and load data from MySQL and PostgreSQL databases, ensuring high data integrity for executive-level reporting. Designed and maintained automated interactive dashboards in Tableau and Power BI, enabling C-suite stakeholders to monitor real-time KPIs and operational efficiency metrics. Implemented Time-Series Forecasting models utilizing Stats-models (ARIMA/SARIMA) to predict inventory requirements, reducing warehouse overstock by 12%. Automated recurring data cleansing and validation workflows using Python scripts and Bash, reducing manual reporting effort by over 20 hours per week and eliminating human entry errors. Integrated and standardized semi-structured data from JSON and XML sources into relational schemas, facilitating more robust cross-functional data analysis and business intelligence. Utilized Scikit-learn for feature selection and engineering, applying Principal Component Analysis (PCA) to reduce dimensionality and improve model training efficiency. Collaborated with business analysts to translate vague commercial requirements into precise technical data specifications and actionable analytical roadmaps. Conducted deep-dive root cause analysis on sales performance dips, utilizing correlation matrices and clustering (K-Means) to segment customer behavior and identify underperforming regions. Established data quality protocols and metadata documentation, ensuring the reliability of enterprise datasets and simplifying the audit process for financial compliance. Performed web scraping using BeautifulSoup to collect competitor pricing data, enriching internal analytics systems for market positioning insights. Prepared and presented weekly technical reports to cross-functional teams, highlighting data-driven recommendations that led to a 5% improvement in overall operational margin. Managed the end-to-end data lifecycle for ad-hoc research projects, from initial data acquisition and cleaning to final model interpretation and visualization. Keywords: cprogramm cplusplus continuous integration continuous deployment quality analyst artificial intelligence machine learning user interface business intelligence sthree active directory rlang microsoft mississippi procedural language New York Texas