Home

Nikhil - Gen AI Engineer

[email protected]

Location: Chicago, Illinois, USA

Relocation: yes

Visa: GC

Resume file: Nikhil_GenAI Engineer_AI_ML_1771620670982.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

Professional Summary 11 years of experience as a Gen AI Engineer, I have designed, built, and implemented analytical and enterprise applications using Machine learning, Python, R, Scala, and Java. Full Stack Engineer with expertise in Voice AI, LLMs, and Cloud Integration, specializing in building AI-driven voice applications and deploying scalable solutions in AWS and vector databases. Familiar with various LLMs, including OpenAI GPT, Claude, Falcon, and Gemini for enterprise GenAI applications. AI/ML Engineer specializing in RAG-based applications, leveraging LLMs (Large Language Models) with vector search, embedding models, and knowledge retrieval to enhance contextual understanding and response accuracy. 2+ years of experience in integrating Large Language Models (LLMs) into voice-driven applications, enhancing conversational AI capabilities for financial and customer support systems. Designed and deployed end-to-end machine learning pipelines using Azure Machine Learning Studio. Utilized AI Artificial Intelligence to allow users to update CRMs such as Salesforce. Designed and implemented an AI Multi-agent System to improve the precision and recall of a web content categorization engine. Strong background in Python-based AI solutions, full-stack development, and integrating telephony, cloud environments, and external APIs for seamless AI interactions. I have extensive experience with machine learning tools and scripts, utilizing platforms like Azure OpenAI, Azure ML Studio, and Prompt Flow, alongside Python and PowerShell scripting. Tested machine learning models applying algorithms based on decision trees, logistic regression, and neural networks. Mentored teams on GenAI + RAG best practices, including vector store selection, embedding optimization, and governance/security. Developed machine learning algorithms and worked with engineering to develop new signals and features. Used Auto-sklearn to automate the machine learning pipeline, reducing the need for extensive domain knowledge and manual experimentation. Automate the complex aspects of the modelling pipeline using auto-learn. Utilized H2O.ai framework that offers high-speed algorithms for machine learning Handled data processing and inference locally on devices, reducing latency by avoiding round trips to the cloud using Edge AI. Acted as a technical SME, providing guidance to cross-functional teams and stakeholders on AI/ML best practices and project deliverables. Enhanced privacy by minimizing the need to transmit sensitive information to the cloud using Edge AI. Used Python Django RESTful framework to create API and Postgres for a database. EDUCATION Bachelors in Electronics and Communication Engineering karunya University (2013) Technical Skills Machine Learning Algorithms: Linear regression, SVM, KNN, Naive Bayes, Logistic Regression, Random Forest, Boosting, K-means clustering, Hierarchical clustering, Collaborative Filtering, Neural Networks, NLP Voice AI & Conversational AI (Speech-to-Text, NLP, LLM-based Voice Assistants) Large Language Models (LLMs) GPT, OpenAI API, LangChain, Fine-tuning Models Analytic Tools Excel, Data Studio Programming Language R 3. X, Python 2. X & 3. X (numpy, scipy, pandas, seaborn, beautiful soup, sci-kit-learn, NLTK), SQL, C Database PostgreSQL, Oracle 11g, MySQL, SQL Server, MongoDB, Neo4j Big-Data Framework Hadoop Ecosystem 2. X (HDFS, MapReduce, Hive 0.11, Hbase 0.9), Spark Framework 2. X (Scala 2. X, Spark SQL, Pyspark, SparkR, Mllib) Data Visualization Tableau 8.0 /9.2 / 10.0, Plotly, R-ggplot2, Python-Matplotlib, Logi Analytics Version Control Git 2.X CI/CD & DevOps Docker, Kubernetes, Terraform, Jenkins Operation System: UNIX, MacOS, Windows Professional Experience Role: Lead GenAI Architect / AI/ML Engineer / LLM Aug 2024 Till Now Client: American Express, AZ Responsibilities: Designed and implemented RAG pipelines integrating LLMs with vector databases (Pinecone, FAISS, Weaviate) to enable context-aware and domain-specific responses. Designed and implemented agentic AI architectures leveraging LLMs, vector databases, and orchestration frameworks to enable autonomous decision-making and workflow execution. Built end-to-end GenAI solutions with RAG for knowledge management, chatbots, and enterprise search, improving accuracy and reducing hallucinations in responses. Integrated transaction-capture devices and edge gateways with cloud ingestion pipelines to collect and transmit encrypted payment and fraud-signal data into Snowflake and Azure analytics platforms. Developed scalable multi-agent LLM workflows using LangGraph, enabling dynamic task routing and memory-aware conversational agents for complex enterprise use cases. Designed and implemented Gen AI solutions leveraging MCP (Model Context Protocol) to enable seamless integration between AI models, developer tools, and enterprise workflows. Optimized prompt engineering and context windowing strategies for RAG to enhance precision and minimize token usage. Integrated IoT-based payment gateways and embedded hardware with NVIDIA-accelerated Jetson and Coral devices to process high-volume transaction data and flag suspicious activity in near real time. Developed custom MCP servers and tools to extend LLM capabilities across IDEs, databases, and internal knowledge bases, accelerating developer productivity. Deployed production-grade RAG architectures on cloud platforms (AWS, Azure, GCP) with scalable APIs and CI/CD pipelines. Built and customized internal AI copilots using Microsoft Copilot Studio to assist customer-support teams with quick policy lookup, transaction summaries, and compliance responses. Integrated Copilot with Azure Cognitive Search and Amex internal APIs to generate real-time, accurate financial data summaries. Assisted in validating inference accuracy and performance between edge devices and cloud-hosted AI models. Applied RAG techniques for multi-modal data (text, PDFs, images, tables) enabling cross-format retrieval and generation. Integrated GenAI models (OpenAI, Anthropic, LLaMA, Mistral) with agent frameworks (LangChain, LangGraph, AutoGen, Haystack Agents) to create domain-specific autonomous agents. Designed and implemented graph-based orchestration for AI agents using LangGraph to manage stateful conversations and decision trees in customer support bots. Implemented edge-side preprocessing on embedded systems using NVIDIA Jetson and Coral to filter, encrypt, and score high-volume card transactions before sending them to Snowflake and Azure for further analysis. Developed RAG (Retrieval-Augmented Generation) pipelines with semantic search, embeddings, and knowledge grounding to enhance agentic AI responses with enterprise data. Utilized Azure Cognitive Services (Computer Vision, Text Analytics, Language Services) to implement intelligent document processing workflows. Designed model orchestration flows using Azure AI Foundry to deploy GPT-based assistants for fraud-risk analysis and dispute resolution workflows. Integrated LangGraph with LangChain, OpenAI GPT-4, and external APIs to build modular AI systems that adapt to user intent across dynamic execution paths. Built custom AI solutions with Azure OpenAI Service, applying GPT-based models for chatbots, summarization, and classification tasks. Optimized prompt engineering and context management within MCP-enabled environments, improving accuracy and relevance of AI-driven outputs. Authored best practices, design patterns, and governance frameworks for scalable and secure deployment of agentic AI systems. Built pipelines where embedded IoT devices and gateways push transaction events, device metrics, and risk signals from Jetson and Coral edge nodes into enterprise data platforms for compliance and reporting. Ensured edge-to-cloud data pipelines were production-grade, secure, and compliant with financial-services regulations. Built Power Automate flows integrated with Amex CRM to trigger AI insights and automate internal service-ticket routing. Created Power Apps dashboards for analysts to visualize transaction risk scores and model recommendations. Delivered secure and compliant AI integrations by implementing access controls, audit trails, and governance mechanisms within MCP workflows. Role: AI Engineer / Full Stack Python Jan 2023 Aug 2024 Client: John Deere, Moline IL Responsibilities Leveraged LangGraph s event-driven architecture to implement long-running agent loops with context preservation and real-time updates in RAG pipelines. Configured Copilot Studio with Azure OpenAI + LangChain backends, providing real-time recommendations for maintenance scheduling. Integrated Big Query ML to build and deploy ML models directly within Big Query using standard SQL, streamlining real-time analytics and predictive modelling. Used IoT sensors connected through industrial gateways and embedded controllers, running NVIDIA Jetson and Google Coral devices, to process real-time telemetry and computer-vision data from agricultural equipment in the field Built and maintained production pipelines that synchronized edge-processed data with cloud platforms for reporting, dashboards, and AI model retraining. Deployed multi-agent workflows combining Azure Machine Learning + Foundry pipelines for continuous model improvement. Integrated IoT and embedded device data from field gateways with NVIDIA-accelerated Jetson and Coral edge platforms to support predictive maintenance and automated decision-making for heavy machinery. Design highly efficient prompts to enhance NLP model accuracy and performance, leveraging vector databases (Weaviate, Pinecone, FAISS) for context memory. Implement MLOps pipelines with MLflow, Airflow, and Docker/Kubernetes to ensure scalable Gen AI deployment in AWS Lambda & Azure Kubernetes Service (AKS). Process large-scale financial data (structured & unstructured) using Apache Spark, Databricks, and Snowflake for AI model training. Designed a hybrid edge-to-cloud architecture where IoT gateways and embedded NVIDIA Jetson and Coral devices performed on-site AI inference while syncing results to Azure and Databricks for analytics and retraining. Created Power BI dashboards powered by Databricks data to visualize equipment failure trends and maintenance KPIs. Deployed ML models as REST APIs using FastAPI and Docker for scalable inference in cloud and edge environments. Worked on DAG-based workflows using Apache Airflow; familiar with concepts applicable to Flink s DAG-based stream processing. Role: (AI/ML Engineer | GenAI | Full Stack Python) Aug 2020 Dec 2022 Client: T-Mobile, TX Responsibilities Created reusable LangGraph nodes for LLM prompting, tool invocation, and branching logic, reducing onboarding time for new workflows by 40%. Build AI-Powered Virtual Assistants Design Gen AI-driven chatbots and voice assistants to improve customer service automation, using LangChain, Rasa, and OpenAI API. Prompt Engineering & Optimization Design, test, and optimize prompts for LLMs to improve response accuracy and context understanding in real-time telecom interactions. Deployed and monitored edge-AI workloads in production across telecom sites, ensuring reliable data flow from gateways to central cloud platforms. Tuned LangGraph workflows for performance and latency, reducing agent orchestration time by 25% while maintaining model accuracy and coherence. Integrated LangChain with OpenAI, Hugging Face Transformers, and custom prompts to build multi-step conversational agents for enterprise document processing. Implemented telemetry and continuous evaluation pipelines inside Foundry to monitor AI performance in production. Implemented logging, alerting, and automated recovery to keep edge and cloud AI systems operating at carrier-grade reliability. Fine-Tune AI Models on T-Mobile Data Train custom AI models on telecom-specific datasets (customer interactions, network logs, and call transcripts) using Hugging Face Transformers, PyTorch, and TensorFlow. Integrated Power Automate with Teams and ServiceNow to automatically route unresolved AI tickets for human review. Used Power BI for visualizing GenAI usage insights, model accuracy, and overall customer satisfaction trends. AI-Driven Speech Recognition & NLP Develop speech-to-text and voice AI solutions using Whisper AI, Kaldi, and Mozilla Deep Speech to improve call canter automation. Environment: TensorFlow, PyTorch, Hugging Face, LangChain, AWS SageMaker, Azure ML, Google Vertex AI, Kubernetes, Docker, MLflow, Apache Airflow, FastAPI, REST APIs. Role: Data Scientist with Full Stack Python Nov 2018 July 2020 Client: Microsoft, WA Responsibilities Aware of client-focused AI or cognitive solutions leveraging Natural Language Processing (NLP), Machine Learning, Probabilistic Decision Making, and related methods and paradigms, and supporting technology. Developed multiple apps using AI and machine learning software. Distribution of computational load across various devices, enabling scalable solutions without overloading centralized systems using edge AI. Transformer architectures Transformer Models have been extended to handle multimodal tasks, such as image captioning, video understanding, and joint processing of text and image data. Used Python to interface with the jQuery UI and manage the storage and deletion of content. Understand search queries and improve search engine ranking and relevancy using BERT. Advanced AI capabilities for natural language processing, content generation, conversation, and various applications using GPT. Provided input representations that capture contextual information, improving the model's understanding of sequences of words using Word Embeddings. Using Big Query, Dataflow, and Apache beam in Google Cloud Platform for end-to-end ML. Environment: AI, Machine Learning, H2O.ai, Edge AI, TensorFlow, PyTorch, Scikit-learn, Keras, MX Net, Neural Networks. Role: Data Scientist with Machine Learning Feb 2016 Nov 2018 Client: Humana, TX Responsibilities Updated an existing real-time bidding platform (RTB) to conform to Open RTB standards. Platform written in Python with Tornado, backed by Redis. Leveraged Miro boards for Agile sprint planning and project tracking, enabling clear alignment on model development stages, compliance checkpoints, and delivery timelines across stakeholders. Developed tools using Python, Shell scripting, and XML to automate some of the menial tasks. Facilitated cross-functional collaboration by using Miro to design and visualize end-to-end ML workflows (ETL pipelines, fraud detection models, and claims prediction systems), improving communication between technical and business teams. Environment: Python, Django, ORM, pandas, Tornado, JavaScript, HTML5, CSS3, Ruby, ROR Ruby on Rails, bootstrap, jQuery, JSON, web token, SSO/SAML, , CVS, SVN, Junit, Waterfall, AWS, EC2, S3, Ant, XML, Jira, Unix, chat. Role: Data Analyst July 2013 Nov 2015 Client: Cigna, India Responsibilities Performed Data Analysis, Data Migration, and Data Preparation useful for Customer Segmentation and Profiling. Architected scalable algorithms using Python programming and capable of performing Data Mining, Predictive Modelling using all kinds of statistical algorithms as required. Developed Multivariate data validation scripts in Python for equity, derivate, currency and commodity related data, thereby improving efficiency of pipeline by 17%. Used Predictive Analysis to develop and design of sample methodologies and analysed data for pricing of client's products. Environment: Python, Pandas, Numpy, Seaborn, Scipy, Matplotlib, Scikit Learn, NLTK, Snowflake, Tableau, jQuery, JavaScript, HTML, NodeJS, Hadoop, MongoDB, OLTP, OLAP, ER Studio, Oracle, SQL Server, SQL, Tableau Server. Keywords: cprogramm continuous integration continuous deployment artificial intelligence machine learning user interface business intelligence sthree rlang Arizona Illinois Texas Washington