Skip to main content
The Nuti family page

  • Gopi Krishna Nuti

    Professional Resume of AI Architect, Author and teacher

    Synopsis

    AI Architect with over 24 years in Data Science and Artificial Intelligence, specializing in Generative AI, Agentic Architectures, and Computer Vision. Designed and built end-to-end RAG and multi-agent systems for the construction industry, from document ingestion and patented chunking through retrieval and production deployment. Holder of 5 patents and author of 4 books, and an active contributor to ML/AI academia and research.

    Core Competencies

    • Generative AI & RAG Systems: Architected end-to-end RAG pipelines — document/PDF extraction, patented chunking algorithms, vector database design (FAISS, ChromaDB) with metadata-driven retrieval, noise elimination, and accuracy tuning. Evaluated and integrated LLMs including GPT-3.5/4, multiple versions of Claude, and AWS Nova Pro.
    • AI Agents & MCP Servers: Designed agentic architectures using LangGraph, LlamaIndex, AutoGen frameworks with sub-agents that query distinct data sources; built MCP Clients and MCP Servers, and investigated and mitigated tool-poisoning vulnerabilities using AWS Guardrails and custom guardrails, with working familiarity with Guardrails.ai and NVIDIA NeMo.
    • Computer Vision: Object detection, localization, classification (YOLO, Inception, ResNet) for face recognition, OCR, and symbol detection; OpenCV/Pillow in C and Python.
    • Machine Learning: Predictive analytics (regression, classification, ranking) across Python, Excel, and SPSS; classical ML and deep learning, with patents in production ML systems.
    • NLP & Conversational AI: Multi-turn, extensible chatbot design; NLTK, AWS Lex, Google Dialogflow; LLM integration.
    • MLOps & Cloud: Built org-wide MLOps framework adopted across Autodesk; AWS (EC2, S3, SageMaker, Athena, Lex, Polly) and Azure ML Studio.
    • Project Leadership: Architecture, tech leadership, hands-on coding, and solution evaluation across cross-geography teams; secondary responsibility for defining AI standards.
    • Languages: Proficient in Python, C, C++, and C#.

    Employment History

     4/2025 - TilldateAI Architect
    Autodesk     		 
    • Owned end-to-end architecture for Autodesk's construction RAG/QnA assistant — from document ingestion through retrieval to response generation — across design, tech leadership, hands-on coding, and solution evaluation.
    • Built the PDF extraction pipeline and a patented document-chunking algorithm for construction specification documents.
    • Designed the vector database schema and retrieval strategy, evaluating and working with FAISS and ChromaDB for chunk storage and metadata-driven retrieval; tuned relevance and reduced retrieval noise.
    • Evaluated and integrated LLMs including GPT-3.5/4, Claude, and AWS Nova Pro, alongside OpenSearch for retrieval, to identify the best fit for accuracy and cost across use cases.
    • Designed the chatbot for extensibility and true multi-turn conversation, and built MCP Clients to connect the chatbot to backend tools and data sources.
    • Architected the agentic backend, building sub-agents that independently query different internal data sources to answer user questions, connected via MCP Servers.
    • Investigated MCP tool-poisoning vulnerabilities and implemented mitigations, combining AWS Guardrails with custom-built guardrails; maintain working familiarity with Guardrails.ai and NVIDIA NeMo for future evaluation.
    • Held secondary responsibility for defining AI standards and best practices for the team.
    • Solution was recognized by customers for meaningfully reducing weekly time spent on document search and supporting better-informed decision-making.
     12/2020 - 3/2025Data Science Manager
    Autodesk

    • Architected patented ML solutions for internal and external customers, spanning Computer Vision, NLP, and Generative AI.
    • Built the initial RAG-based chatbot for Autodesk Construction Cloud using OpenSearch and LLMs, and the early agentic/MCP Server architecture underlying it.
    • Delivered Computer Vision solutions for architectural symbol detection (Takeoff workflow) using OpenCV, YOLO, Inception, and ResNet — reducing manual effort by 90%.
    • Designed and built a state-of-the-art MLOps framework that reduced time-to-market for ML solutions, later adopted org-wide.
    • Drove technical direction for AI adoption across the organization, aligned to policy and best practices.
       
     03/2020 – 11/2020
    Lead Data Scientist,
    Enterprise Startup 
    • Architect ML solutions for addressing business cases in the areas of forecasting, Chatbots, behavioural modelling etc using Time series analysis algorithms, NLP, predictive analytics and AutoML.
    • Upgrade the ML pipeline to state-of-the-art technologies to ensure the competitiveness of the product. 
    • Achieved an improvement 20% in forecasting accuracy. 
    • Achieved classification accuracy of 96% in behavioural modelling for customers. 
    • Built chatbots which have attracted venture capitalist interest 
     07/2002 – 02/2020
     Senior Data Scientist,
    Tata Consultancy Services Ltd.
    • End to end ownership for data science projects.
    • Design and develop methods and solutions to analyse structured and unstructured data sources.
    • Achieved an OpEx improvement of $2.5 Mn for the customer.
    • Team has won the Edison award for innovation in 2012.
    • Winner of the “Most Valued Player” award from Vice President.
    • Won multiple awards for contribution.  

    Patents

     1 
    Generative Artificial Intelligence (AI) Construction Specification Interface
    		 
    US Patent 12,596,732-B2
    issued April 7, 2026
     2 
    Automatically Sectioning of Construction Specification Documents for Query Optimization
    		 
    US Publication 2025/0298819-A1,
    published September 25, 2025
     3 
    A System and Method for Personalized Query and Interaction Set Generation Using Natural Language Processing Techniques
    		 
    USPTO Application No. US 17/082,842
     4 
    A System and Method for Automatic Update of Customer Relationship Management and Enterprise Resource Planning Fields with Next Best Actions Using Machine Learning
    		 
    USPTO Application No. US 17/082,879.
     5 
    A System and Method for Automatic Ranking of Opportunities in Sales and Revenue Forecasting Using Machine Learning
    		 
    Indian Patent Application No. 202141050724
     6 
    Provisional patent declarations
    		 30566.0619USU1,
    30566.0623USP1.
    1.
    Generative Artificial Intelligence (AI) Construction Specification Interface
     US Patent 12,596,732-B2, issued April 7, 2026.
     
    1. Automatically Sectioning of Construction Specification Documents for Query Optimization
    US Publication 2025/0298819-A1, published September 25, 2025.

    2. 
    A SYSTEM AND METHOD FOR PERSONALISED QUERY AND INTERACTION SET GENERATION USING NATURAL LANGUAGE PROCESSING TECHNIQUES. 

    US PTO Application US 17/082842. 

     3.
    A SYSTEM AND METHOD FOR AUTOMATIC UPDATE OF CUSTOMER RELATIONSHIP MANAGEMENT AND ENTERPRISE RESOURCE PLANNING FIELDS WITH NEXT BEST ACTIONS USING MACHINE LEARNING 

    		US PTO Application  US 17/082879  
     4

    A SYSTEM AND METHOD FOR AUTOMATIC RANKING OF OPPORTUNITIES IN SALES AND REVENUE FORECASTING USING MACHINE LEARNING 

    Indian PTO application 202141050724 

     5. Provisional patent declarations30566.0619USU1 
    30566.0623USP1 

    Publications

    1. 

    CompTIA Cloud+ Certification Guide (Exam CV0-004)   

    		ISBN-13: 978-9365890594 by BPB Publishers 
    2.Neural Network Computer Vision with OpenCV 5 ISBN-13: 978-9355516961 by BPB Publishers  
     3.CompTIA Cloud+ Certification Guide (Exam CV0-003)  ISBN-13: 978-9355513847 by BPB Publishers 
     4.Machine Learning for Engineers  ISBN-13: 978-9389024876 by Penman Publishers 
     5.Using Data drift to identify Concept drift in Machine learning models in International Journal of Computer Engineering and Applications · Nov 2020 

    Certifications

    Specialist - SPSS Statistics Level 1 v2 IBM Certified 
    EdX Certified R developer Microsoft 
    Cloud Essentials  CompTIA
    Cloud Architect CompTIA 
     PMP  (expired)Project Management Institute

    Select Projects

      Chatbot for Construction Specifications
      QnA Solution using Generative AI Developed an innovative Question and Answer (QnA) assistant specifically tailored for the construction industry. The solution leverages advanced technologies including OpenSearch for realtime data retrieval, RAG (Retrieval-Augmented Generation) for integrating retrieved documents into a unified response, and LLMs for enhancing natural language understanding and response generation. This allowed for a highly effective and user-friendly tool that significantly outperformed competitor solutions in terms of accuracy and user satisfaction. 
      Built an Agentic solution architecture for the chatbot with MCP Clients in the chatbot and MCP servers on the backend. Responsible for identifying security loop-holes like tool poisoning in the MCP Servers and figuring out ways to fix them. 


      Symbol Detection in floor plans for Takeoff functionality 
      Conceptualized and implemented a robust Computer Vision solution for construction industry tailored to detect both custom and standard symbols on architectural floor plans for the 'Takeoff' workflow. The system was built using advanced technologies such as OpenCV for image processing, alongside state-of-the-art deep learning models including YOLO, Inception, and ResNet for symbol detection. This groundbreaking solution has streamlined the customer's workflow, reducing their effort by nearly 90%, and significantly improving efficiency and productivity.  
      ​

      Bottom-up Revenue forecasting for B2B organizations 
      Developed and patented a machine learning system for sales organizations to enhance their monthly revenue forecasting. This project was accomplished through the application of various time series analysis techniques on Salesforce data, which facilitated the construction of predictive analytics models such as regression, classification, and ranking. The system's accuracy was significantly improved from 65% to an impressive 92% through rigorous work on MLOps and Machine Learning architectures, thereby elevating the competitive edge of the system. Technologies utilized in this project include MongoDB, Time Series Analysis, ARIMA, and Ranking algorithms.  
      ​

      Intelligent Voice Assistant (Chatbot) 
      Engineered a patented context-aware chatbot, designed to proficiently handle queries related to the sales process. This project involved extensive use of Natural Language Processing (NLP) and Natural Language Understanding (NLU) techniques, along with advanced AWS services like Lex, Polly, and Comprehend, as well as Google Streaming Transcription. The patented product not only delivers context-sensitive responses, but also proactively provides intelligent insights drawn from internal databases. This empowers sales representatives to deliver informed and effective responses.  
      ​

      Diabetic Retinopathy Detection using Smartphone 
      This project involves building a Machine Learning model to detect symptoms of Diabetic Retinopathy for a person. InceptionV3 model was used for classification. Custom algorithm was built to pre-process the images fed to InceptionV3. The model achieved an accuracy of 84.5% on a curated Kaggle dataset and a test accuracy of 78% on the Messidor dataset. This is an improvement of ~28% compared to out-of-the-box Inception V3’s accuracy. The model was then ported to Qualcomm Snapdragon’s SNPE format and using a mobile app, was run on the mobile phone (OpenCV, Keras, Inception V3, Python). 
      ​

      Smart Defect detection in Manufacturing
      This project involves performing video analytics to recognize manufacturing defects in PCB assembling. Architected and built the Machine Learning pipeline, MLOps and production process using on-prem servers. (OpenCV, Dlib, YOLO, Tesseract, Python). 
      ​

      Customer Recognition and Identification 
        This project involves performing video analytics and extracting insights for Retail customers. Customers had to be uniquely identified and tracked without extensive training data. The activities done were Face Recognition, Object recognition and classification, footfall analysis and Gesture recognition. Made extensive use of OpenCV, Dlib, Python and software algorithms like YOLO and OpenFace (OpenCV, Dlib, OpenFace, CNNs, YOLO). 


        Television Audience Measurement and Prediction 
        The requirement is for measuring and predicting Television audience. Audience count for individual programs is collected daily in structured and semi-structured formats. Using this data, the number of possible audiences for next day is predicted. This solution won the Edison award for the year 2012.

        Volunteering

        2018-06 - present MUST Research Club 

        MUST Research Club is a technology NGO working to reduce the gap between Industry and Academia in the field of Machine Learning and Artificial Intelligence. As it's Vice President, I worked extensively with state government of Telangana, industry bodies like NASSCOM, BIS and with various engineering colleges in Telangana and Andhra Pradesh. Conducted multiple workshops and training sessions for students. Participated in discussions with State Government of Telangana on the Emerging Technologies thought leadership roadmap for the government. I volunteered on below listed projects. 

        • Advertising Brand Exposure Measurement using Computer Vision Process video streams in a TV/Movie to measure the visibility of a brand logo. This is used for measuring the advertisement effectiveness of different spots in a stadium or on players' uniforms. (OpenCV, Resnet, YOLO, InceptionV3, Python) 
        • Face Recognition for a Robotic Personal Assistant Developed face recognition extension for a robotic personal assistant. This will allow the personal assistant to have a customized response for each individual member of the household. (OpenCV, Openface, Pytorch, Python) 
        • Implementation of a perceptron Implemented a perceptron with forward propagation, backward propagation and activation functions using C in a platform agnostic way. This project was intended to provide an easy understanding of perceptron’s behaviour and compare performance on a variety of hardware.

        Industry events

        • Member of board of studies for CSE department in Geetanjali Institute of Science and Technology
        • Industry speaker at multiple events 
          • Special invitee from industry at NatFoE-2023 at BIT, Mesra 
          • Mathematics, ML and ML Engineering at RAMSA - 2021 by GVP College of Engineering
          • AI for Pharmaceutical Industry at AU Engineering College 
          • Classification-vs-Ranking at Srinidhi College of Engineering
          • Faculty Development Program at Mallareddy College of Engineering 
          • Emerging technologies and soft skills at F2 event, Vijayawada
          • Edge Analytics at NASSCOM’s BRAINS event, Hyderabad 
          • Keynote speaker at International Conference on Emerging Trends in Information Technology, Oct 2020 
          • Guest lecturer in multiple engineering colleges in AP and Telangana on Machine Learning. 
          • Technical Speaker on Computer Vision during Jury member for various hackathons on AI/ML

        Blogs and Social Media

         Github https://github.com/ngopikrishna 
         Mediumhttps://medium.com/@ngopikrishna.public 
          Codeprojecthttps://www.codeproject.com/Articles/Gopi-Krishna-Nuti#Article  
         Slidesharehttps://www.slideshare.net/gopikrishnanuti/ 
          
      Gopikrishna Nuti //