Architected enterprise-grade LLM-based predictive maintenance platforms by integrating GPT-4, GPT-3.5, and LLaMA models with Gong, Salesforce, and N8N to automate maintenance scheduling and failure diagnosis.

Developed LangChain-powered RAG agents to provide multi-turn contextual aerospace support, including document summarization, equipment troubleshooting, and federated search across Salesforce Knowledge and Service Cloud.

Integrated Outreach and Salesforce Einstein APIs with ChatGPT and Claude to design AI-driven agent workflows for sales enablement, contact prioritization, and sentiment-aware escalation.

Orchestrated asynchronous A2A (Agent-to-Agent) communication patterns using LangChain multi-agent routing with task-specific agents for diagnostics, telemetry summarization, and replacement part recommendation.

Built N8N automation workflows for event-triggered LLM tasks such as maintenance report generation, document enrichment, and Slack-based approvals.

Deployed enterprise-compliant GenAI microservices using FastAPI, Kubernetes (EKS), and Docker with routing logic for Claude, ChatGPT, and private LLaMA-based endpoints.

Developed retrieval-augmented chat agents backed by Pinecone and FAISS for real-time question-answering over engineering manuals and cloud-stored telemetry, reducing lookup latency by 40%.

Enabled cross-cloud AI workloads using Azure Cognitive Services for ingestion and LLM fusion from AWS and Azure, enabling hybrid document retrieval and summarization.

Implemented Reinforcement Learning from Human Feedback (RLHF) pipelines to iteratively improve agent reasoning in maintenance and procurement workflows.

Designed telemetry dashboards using Grafana and Prometheus to track agent token usage, hallucination rate, and routing logic accuracy across enterprise AI platforms.

Led governance initiatives on AI explainability and prompt versioning using LangSmith, MLflow, and GitHub Actions, ensuring compliance in highly regulated aerospace environments.

Mentored teams on agent orchestration, prompt engineering for Salesforce/Gong/N8N contexts, and LLMOps practices including drift detection and rollback strategies.

Led the architecture and deployment of HIPAA-compliant healthcare chatbots integrating ChatGPT APIs, Salesforce Health Cloud, and Outreach automation to assist clinicians in diagnosis support and EHR retrieval.

Led the end-to-end design and deployment of a HIPAA-compliant GenAI assistant using GPT-4, GPT-3.5, and fine-tuned LLaMA-2/GPT-J models.

Built high-performance RAG pipelines (LangChain, FAISS, Pinecone, OpenSearch) for real-time clinical Q&A and EHR search.

Integrated multi-agent orchestration to route queries between OpenAI, Claude, Cohere, and private LLaMA endpoints based on sensitivity.

Developed multi-modal healthcare chatbots with Slack, Teams, and AWS ECS integrations, supporting diagnosis, summarization, and medication checks.

Implemented FastAPI + Docker + Kubernetes deployments for scalable, secure API access to GenAI services.

Automated PDF ingestion (500K+) using Apache Tika + Haystack, with custom chunking for scalable document retrieval.

Built clinician-facing dashboards (Streamlit, Power BI) for real-time visibility into usage, performance, and drift metrics.

Ensured LLMOps maturity with MLflow, LangSmith, and DVC for version control, model lineage, and reproducibility.

Enabled secure vector search using Sentence Transformers + Pinecone/OpenSearch, reducing retrieval latency by 50%+.

Mentored engineers and analysts on prompt engineering, LangChain agents, vector optimization, and GenAI best practices.

Defined and enforced AI/ML best practices aligned with internal governance and federal compliance frameworks (HIPAA, GDPR, NIST).

Formulated predictive models to forecast product category-wise order volumes and season-wise color and style choices for departmental buyers to make educated and data-driven decisions using Python and/or PySpark.

Implemented time series forecasting models (ARIMA) to predict trends of fuel consumptions for different flight engines.

Worked on image classification using CNN and Computer Vision and implemented hyperparameter tuning for scaling performance, achieving over 87% accuracy.

Applied ML algorithms and statistical modeling techniques like Decision Trees, Naive Bayes, Principal Component Analysis, regression models, Artificial Neural Network, clustering, and SVM to identify volume using Scikit-Learn packages in Python/PySpark.

Developed keyword extraction models using TF-IDF, Word2Vec, NLTK, and other NLP packages.

Implemented parallelized data processing operations using Dask framework to clean and filter text data using Python and/or PySpark.

Responsible for SSRS planning, architecture, training, support, and administration in development, test, and production environments.

Train ML models on large datasets using SageMaker’s training capabilities.

Applied AI/ML algorithms and statistical modeling like Decision Trees, text analytics, image and text recognition using OCR tools like NLP, supervised and unsupervised, regression models.

Wrote Python routines to log into the websites and fetch data for selected options, used Python modules of urllib, urllib2, requests for web crawling, and ML techniques: clustering, regression, classification, graphical models.

Used GCP, BigQuery, GCS Bucket, G-Cloud Functions, Cloud Dataflow, Pub/Sub Cloud Shell, GSUTIL, BQ command line utilities, Data Proc, and Stackdriver and worked on Confluence/Jira and data visualization like Matplotlib and Seaborn library.

Worked with fact dimensional modeling (Star Schema, Snowflake schema), transactional modeling, and SCD (slowly changing dimension), process and load bound and unbound data from Google Pub/Subtopic to BigQuery using cloud Dataflow with Python.

Developed different statistical Machine Learning and data mining solutions to various business problems and generated data visualizations using R, Python, and Tableau.

Involved in the development of web services using SOAP for sending and getting data from the external interface in XML format.

Worked on the development of SQL stored procedures on MySQL, reduced code redundancy to the optimal level, and designed and built a text classification application using different text classification models.

Built and architected multiple data pipelines and end-to-end ETL and ELT processes for data ingestion and transformation in GCP and used AWS components like EC2 and S3.

Performed data analysis, data migration, data cleansing, transformation, integration, data import, and data export through Python.

Developed and deployed data pipelines in cloud such as AWS and GCP.

Devised PL/SQL stored procedures, functions, triggers, views, and packages.

Implemented Apache Airflow for authoring, scheduling, and monitoring data pipelines.

Analyzed large datasets from manufacturing operations, warranty claims, and equipment performance logs to uncover inefficiencies and recurring issues.

Created dynamic dashboards and interactive reports using Tableau and Power BI to visualize KPIs for business units and executive teams.

Designed and maintained SQL queries for data extraction, transformation, and loading (ETL) from Oracle, SAP, and legacy databases.

Conducted statistical analyses (e.g., ANOVA, regression, control charts) to support product quality improvements and root cause investigations.

Automated recurring reports and data processing tasks using Excel VBA and macros, saving 10+ hours per week in manual effort.

Collaborated with product design and engineering teams to translate performance data into actionable product enhancement recommendations.

Supported supply chain optimization by analyzing vendor performance metrics and forecasting material needs using historical trends.

Led ad hoc data projects to support marketing, sales, and customer support initiatives with insights on usage behavior and regional demand.

Conducted time series and trend analysis on machine telemetry data to anticipate service needs and reduce field failures.