AI/ML Engineer Generative AI & LLMOps Specialist

Pavankalyan Dasari

+1-217-***-**** *********************@*****.*** linkedin.com/in/pavankalyandasari

Professional Summary:

Innovative and results-driven AI/ML Engineer with nearly six years of experience building enterprise-grade Generative AI, LLMOps, and NLP systems across cloud-native, high-volume environments. Skilled in designing end-to-end RAG architectures, vector database pipelines (FAISS, Pinecone, ChromaDB), ingestion frameworks, hybrid search, and retrieval APIs using LangChain, LangGraph, and LlamaIndex. Adept at developing backend AI services, integrating LLM APIs, and deploying reliable, observable, and production-ready systems on AWS, GCP, Docker, and Kubernetes. Strong background in implementing monitoring, logging, and performance metrics to ensure transparency, safety, and high availability of AI-driven workflows.

Known for collaborating with cross-functional teams to build prototypes, experiment with emerging AI tools, and deliver human-centered, impact-focused solutions. Experienced in semantic search, multi-agent orchestration, compliance-aware AI practices, and secure data pipelines that align with responsible AI and data governance standards. Passionate about shaping modern AI infrastructure for real-world applications and enabling organizations—including public-sector teams—to adopt scalable, ethical, and efficient AI capabilities.

Category

Tools & Technologies

AI & Machine Learning

GPT-3.5, GPT-4, LLaMA-2, LangChain, LangGraph, AutoGen, Hugging Face Transformers, OpenAI API, Prompt Engineering, RAG, Multi-Agent Systems, Semantic Search, Reinforcement Learning (RL), TensorFlow, PyTorch, Keras, Scikit-learn, Explainable AI (XAI), SHAP.

Natural Language Processing

SpaCy, NLTK, sentence-transformers, LangChain Agents, NER, Text Summarization, Document Understanding, QA Systems, Prompt Orchestration, OCR Parsing.

AI Agents & Orchestration

LangGraph, AutoGen, CrewAI, MCP Protocol, Agent Tooling, Multi-Agent Collaboration, Long-Horizon Reasoning, Planning Algorithms, Agentic Workflows, LLM-as-Judge Evaluation, Knowledge Retrieval, Agent Memory Systems.

Data Engineering & Databases

Python (Pandas, NumPy), SQL, Apache Airflow, Glue Catalog, Athena, PySpark, FAISS, Pinecone, ChromaDB, Weaviate, DynamoDB, MS SQL Server, SSIS.

Cloud & DevOps

AWS (Lambda, SageMaker, EKS, S3, EC2, CodePipeline, CodeBuild), Google Cloud (Vertex AI, Cloud Functions), Terraform, Docker, Kubernetes, Jenkins, Git, CI/CD Pipelines, Model Monitoring, Vector Store Indexing.

Visualization & Reporting

Power BI, Power BI Report Builder, Power Automate, Tableau, Drill-through Dashboards, DAX.

Computer Vision

OpenCV, TensorFlow, Keras, PyTorch, Image Classification, Object Detection, Document Parsing, OCR Pipelines.

Educational Details:

University Of Illinois, Springfield, Illinois Aug 23 - May 2025

●Master of Science (M.S.): Management Information Systems.

Jawaharlal Nehru Technological University, Hyderabad, India Aug 2015 - May 2019

●Bachelor of Technology: Aeronautical Engineering.

Professional Experience:

ABM Industries Wheeling, IL

Gen AI Developer Feb 2024 – Current

Roles & Responsibilities:

●Designed and deployed advanced AI solutions leveraging LLMs GPT-3.5, GPT-4, and HuggingFace Transformers to enable real-time summarization, semantic search, and intelligent document understanding for enterprise-grade applications.

●Implemented Retrieval-Augmented Generation (RAG) frameworks with real-time data feeds to power LLM-backed virtual assistants capable of sub-second response times for anomaly detection, technical support, and compliance tracking.

●Integrated transformer-based models with LangChain, LangChain Agents, and LangGraph to build context-aware AI agents capable of dynamic reasoning, multi-step planning, and tool-assisted task completion in production systems.

●Developed multi-agent orchestration workflows where reasoning agents, retrieval agents, and execution agents collaborate autonomously using frameworks like AutogenAI, MCP Protocol, and LangChain Agent Executors.

●Integrated transformer-based models with LangChain to build context-aware AI agents capable of dynamic reasoning over domain-specific datasets, enhancing interpretability and knowledge retrieval in production environments.

●Led the development of semantic search systems combining vector embeddings and metadata filtering to support fast, high-precision retrieval from large-scale knowledge bases and technical repositories.

●Built robust data ingestion and preprocessing pipelines for time-series sensor data and document corpora, applying NLP techniques including text normalization, sentence segmentation, and entity recognition to optimize input for transformer models.

●Architected and fine-tuned multi-model NLP pipelines to transform unstructured sensor logs and installation documents into structured, context-rich summaries, significantly improving operational decision-making and user experience.

●Designed scalable, containerized AI microservices for real-time inference, orchestrated using Kubernetes and serverless workflows, ensuring high availability and minimal latency for model-backed applications.

●Deployed and managed AI inference workloads on AWS, using services such as AWS Lambda, Amazon SageMaker, and Amazon EKS as supporting infrastructure to ensure scalability, security, and cost-efficiency of AI systems.

●Collaborated with cross-functional teams to embed AI capabilities into user-facing applications, including intelligent chatbots, real-time monitoring dashboards, and smart documentation interfaces.

●Instrumented observability layers into AI systems to capture inference latency, model drift, and input anomalies, feeding back into iterative model retraining and reliability optimization.

●Spearheaded experiments in prompt engineering, few-shot learning, and embedding tuning to maximize the responsiveness, reasoning, and context handling of deployed LLMs in real-world scenarios.

●Contributed to the continuous evolution of enterprise AI strategy by identifying high-impact use cases, translating business needs into AI solutions, and guiding adoption of generative AI technologies at scale.

Tech Stack: Python, HuggingFace Transformers, LangChain, GPT-3.5 / GPT-4, PyTorch, TensorFlow, Kubernetes, Docker, AWS (Lambda, SageMaker, EKS, S3), REST APIs, RAG architectures, Faiss, OpenSearch, ElasticSearch, Pinecone, Prompt Engineering, Semantic Search, Vector Embeddings, PySpark, Airflow, Git, CI/CD (GitHub Actions / CodePipeline), JSON, YAML, AutogenAI, MCP Protocol, LangChain Agent Executors, context-aware AI agents.

AIR INDIA SATS Pvt Ltd Hyderabad, India

Turnaround Coordinator (AI/ML/Data Scientist) Aug 2019 – Sept 2022

Responsibilities:

Developed end-to-end NLP pipelines for document classification, summarization, and information extraction using spaCy, NLTK, TF-IDF, Word2Vec, GloVe, FastText, and early transformer models such as BERT, DistilBERT, and T5 streamlining aviation documentation workflows for Ground operations.

Built predictive ML models using Random Forest, SVM, XGBoost, Logistic Regression, and K-Means for delay prediction, workload forecasting, risk scoring, and aircraft turnaround optimization.

Designed and implemented OCR-based automation systems with OpenCV and Tesseract OCR to digitize load sheets, crew logs, and equipment checklists reducing manual effort and improving data accuracy.

Engineered deep learning models including CNN, LSTM, and BiLSTM for image-based equipment detection, crew activity analysis, and anomaly detection in aviation operations.

Built and maintained ETL pipelines using Apache Airflow, Python (Pandas, NumPy, SciPy) and SQL to process structured and unstructured operational data from flight systems and ground-handling logs.

Developed REST-based microservices using Flask and FastAPI to expose ML predictions to Emirates dashboards, enabling real-time decision support during aircraft turnaround.

Deployed scalable ML workloads using AWS SageMaker, EC2, S3, and Lambda, enabling efficient model training, batch processing, and serverless inference.

Implemented CI/CD workflows with Jenkins, GitHub Actions, Docker, and Kubernetes for automated deployment, versioning, testing, and rollback of ML applications.

Performed advanced statistical modeling and feature engineering techniques including PCA, time-series forecasting (ARIMA/LSTM), and anomaly detection, improving prediction accuracy and operational reliability.

Integrated ML outputs with BI dashboards using Power BI and custom APIs to provide real-time insights into aircraft readiness, GSE (Ground Support Equipment) movement, and operational SLAs.

Established monitoring pipelines to track model drift, inference performance, and data quality using CloudWatch and custom monitoring scripts in Python and Bash.

Automated extraction of operational KPIs such as load sheet data, fuel estimates, and equipment timestamps, reducing manual errors and improving decision-making for ground-handling teams.

Collaborated with Emirates operations, engineering teams, and aviation safety teams to validate ML models, ensuring accuracy, compliance, and integration into critical airport workflows.

Tech Stack: Python (Pandas, NumPy, SciPy), Scikit-learn, TensorFlow, PyTorch, spaCy, NLTK, TF-IDF, Word2Vec, FastText, GloVe, Sentence Embeddings, Transformer models (BERT, DistilBERT, T5), OpenCV, Tesseract OCR, Apache Airflow, SQL (MySQL, PostgreSQL), Flask, FastAPI, Docker, Kubernetes, Jenkins, GitHub Actions, AWS (SageMaker, EC2, S3, Lambda, CloudWatch), Linux, Bash scripting, and classical ML algorithms (Random Forest, SVM, XGBoost, K-Means, Logistic Regression).

HDFC ERGO Hyderabad, India

Data Analyst/Scientist Dec 2018 – Jul 2019

Roles & Responsibilities:

●Designed and developed interactive Power BI dashboards with drill-through reports and custom visuals, improving user experience and enabling deeper business insights.

●Optimized data models using DAX expressions to enhance report performance, reduce load times, and increase efficiency in business analysis.

●Integrated data from multiple sources, including SQL Server databases and Azure Data Lake, to create unified datasets for comprehensive reporting.

●Developed ETL pipelines using SQL Server Integration Services (SSIS) and Azure Data Factory to extract, transform, and load large datasets, ensuring timely and accurate data availability.

●Performed data cleaning, profiling, and feature engineering using Python libraries such as Pandas and NumPy to prepare data for machine learning and analysis.

●Built machine learning models including Support Vector Machines, Random Forest, and XGBoost using Python’s Scikit-learn and PySpark MLlib to predict customer behavior and improve decision-making.

●Deployed and managed machine learning models in production environments using Amazon SageMaker, with CI/CD automation via AWS CodePipeline and CodeBuild for continuous integration and delivery.

●Automated report scheduling and alerting through Power Automate and Power BI Service and created paginated reports using Power BI Report Builder to accelerate report generation and distribution.

Tech Stack: Power BI (DAX, Report Builder, Power Automate), SQL Server, Azure Data Lake, SSIS, Azure Data Factory, Python (Pandas, NumPy, Scikit-learn, PySpark MLlib), Amazon SageMaker, AWS CodePipeline, and CodeBuild.

Notable Academic Projects:

Offline Agentic RAG System (FAISS + LangChain + FastAPI)

Built a fully offline, multi-document Agentic RAG system using FAISS, LangChain Core, FastAPI, and Ollama (Llama 3.1) to enable private local retrieval-augmented generation without cloud dependencies.

Engineered dual pipelines Basic RAG and Agentic RAG where the agentic workflow used tool-calling, reasoning, and multi-step context synthesis, yielding 10 more accurate and structured responses than standard retrieval.

Designed modular multi–vector-database architecture supporting FAISS (offline) and Pinecone (future online mode), enabling scalable document ingestion, semantic search, and hybrid RAG experimentation.

Global Cybersecurity Threat Analysis

●Achieved 90% accuracy in forecasting and detecting cybersecurity threats using SAS Enterprise Miner on global datasets within a Citrix VDI environment.

●Designed Tableau dashboards to visualize threat trends and strengthen security protocols, reducing potential system risks by 15%.

Surrogate CFD Model for 2D Airfoil

Built a geometry-aware surrogate AI model using PyTorch MLPs to emulate 2D aerodynamic CFD simulations (R = 0.914, MSE = 0.0469), enabling real-time prediction and design optimization.

Automated the CFD-to-ML workflow with autonomous preprocessing, training, and validation agents—advancing physics-informed, data-efficient modeling for energy and aerospace systems.

Chicago Education Advocacy Cooperative (ChEAC) — Data Science Fellow

Developed an AI-driven project-matching assistant that parsed resumes, extracted skills, and matched volunteers to initiatives via LLMs, NLP, and vector-similarity search.

Led AI literacy and ethics modules promoting responsible, sustainable technology adoption—bridging community education with agentic and socially-aware AI practices.

Awards & Certifications

●Oracle Cloud Infrastructure 2025 Certified AI Foundations Associate, Oracle, 2025

(OCI AI Services, ML Fundamentals, Generative AI Basics, Responsible AI)

●Neo4j Certified Professional – Graph Databases & Cypher, Neo4j Graph Academy, 2025

(Graph Data Modeling, Cypher Query Language, Full-Text & Vector Indexes, Query Optimization)

●ARIS Business Process Analysis Platform, 2023 – University of Illinois at Springfield

●SQL Essential Training, 2023 - National Association of State Boards of Accountancy, NASBA,2024

●CATIA V5 Associate - Part Design, Dassault Systems, 2019 (Certificate ID: C-QX9MT8D2EE)

●Certified Machine Learning Engineer – Udemy

●RC Aircraft Design & Demonstration - Technozion ’18, National Institute of Technology, Warangal

Contact this candidate