Data Scientists, Machine Learning, Statistics

Dedicated and Experienced professional specializing in Machine Learning, Generative AI, and Data Science, with Ph. D in AI technologies. Currently working with Adobe as a part of GEN AI Center of Excellence team as AI Engineer. Seeking opportunities in the organization to contribute expertise and drive innovation in advance AI solutions. CORE EXPERTISE & TECHNICAL SKILLS

• Generative AI & Machine Learning: Hands on experience in Generative AI frameworks (Langchain, Langflow,CrewAI, LangGraph, Llama-Index), Prompt Engineering (Zero-shot, Few-shot, COT), MCP, A2A, Advanced Retrieval-Augmented Generation (RAG), Multi- Agentic system and applications with Large Language Model(LLM) .Foundation model-based agents, Agentic evaluation and tooling, Ranking Models, Sentence transformer, Experience analyzing data and training model with various machine learning algorithms, Hybrid Models and Neuro-Symbolic AI, Vision Language Models(VLM)

• Natural Language Processing (NLP): Expertise in text processing, topic modeling, and information and entity extraction using BERT, NLU and NLG using LLM, Text Embedding Models, Adobe Firefall, open- source models using Hugging face and frameworks like SpaCy, Pyspark, NLTK, Fast text, TextBlob.

• Deep Learning & Image Processing: Hands on experience in fine-tuning Language Models using frameworks like TensorFlow, PyTorch, along with Lora/Qlora techniques. Contrastive Learning for Computer Vision, Stable Diffusion models (v2.1,3,3.5 medium, 3.5 large, VAE, Text encoders, Transformers, CLIP, T5XXL), Dreambooth, Automatic1111(Text-to-Image, Image-to-Image,Text-to- Video), Text-to-Audio, Audio-to-Text, image/object recognition, Video Generation, segmentation, evaluating AI models or systems.

• MLOps/LLMOps: Hands-on experience with Terraform (IaC), Kubeflow (K-Native), Azure Databricks, Snowflake, and MLflow for experiment tracking, model registry, and batch model serving. Proficient in deploying AI solutions using Docker, Kubernetes, and Flask/ Fast REST API on AWS, GCP,Azure clouds.

• Tools & Platforms: IBM WatsonX, Adobe Firefly, Adobe Mix Modeler (MMM), GitHub, Airflow, Celonis, Jupyter Notebook, Google Colab, VS Code, Postman, Elastic Search, OpenSearch, Ollama, Neo4j.

• Vector Database: Experience with both structured and unstructured data, Expertise in FAISS, Pinecone, CosmosDB, Adobe Emerald and other vector-based database technologies. Utilized vector databases for similarity search, embeddings management, and Generative AI applications.

• Research & Mentorship: Published various research papers on ML, NLP, and SLM/LLM optimization in international conferences and Journals. Actively mentored teams in AI, NLP, advanced ML concepts and Marketing Mix Modeler using AMM within Adobe.

• Industry Applications: Implemented AI and ML solutions in Retail, T&H, CPG and Automotive industries across NA, APAC, and European regions.

PROFESSIONAL EXPERIENCE

ADOBE

Awarded with the "Big Bet Award" for outstanding contributions as a member of the GenAI Center of Excellence (COE) team at Adobe for Q3, 2024. Senior AI ML Platform Engineer Senior Data Scientist (Oct 2023 – Present) Contributing to Gen AI Platform and frameworks development which provides a comprehensive ecosystem for building multi-agentic systems, offering interfaces for LLMs, tools, and orchestration through a graph-based approach. The low-code platform enables users to define agents, prompts, and tools using a drag-and-drop interface, generating deployable Kubernetes-based code. It seamlessly integrates with Adobe's ecosystem (Commerce, AEP, AEM) and supports multiple LLMs, including Adobe's proprietary Firefly model. The architecture supports flexible deployment across major cloud platforms like AWS and Azure, providing scalability and versatility. Extensive knowledge and experience with Marketing Mix Modelling & Adobe Mix Modeler to provide data driven insights, train models and create spend plans.

Currently spearheading the development of an AI-powered Search, Discovery, and Content Intelligence system for hyper-personalization by designing a Hybrid Search framework that integrates Knowledge Graphs (Graph RAG, Neo4j), Semantic Search (LLM, Sentence Transformers), and Keyword Search (BM25). Enhancing search relevance and ranking through Cross Encoders, ColBERT, and Reinforcement Learning-based models, while ensuring seamless integration with Adobe Experience Platform (AEP) data and Vector Databases to optimize retrieval efficiency and chat like user experience.

Developed a virtual try-on feature by developing a sophisticated multimodal vision model using cutting-edge Generative AI. The project leverages computer vision techniques along with finetuned Stable Diffusion 3.5 Medium with complex architectural components like CLIP and Text Encoders, T5XXL transformers, advanced schedulers, VAE, and Unet to transform product visualization. By implementing contrastive learning techniques and utilizing DreamBooth for hyper-personalized model fine-tuning, I'm applying Lora methodologies to optimize model performance efficiently. The entire process is powered by CUDA-accelerated training across Python libraries including PyTorch, Accelerator, Diffusers, xformers and Triton, enabling precise image rendering and reconstruction using custom product image datasets for e-commerce industry.

Developed a sophisticated custom validation model by finetuning DeBERTa v3 base model using Lora, peft with 99% accuracy designed to proactively identify and mitigate potential security risks in AI interactions, specifically targeting prompt injections, potential jailbreak attempts, and distinguishing between benign and malicious user queries. By meticulously training this model and integrating it as a specialized assistant(neural) agent within the concierge layer of our multi-agentic application in production using kubeflow on AWS cloud.

Established a robust Responsible AI and custom Guardrail model that dynamically assesses and filters user interactions with hybrid approach along with NeMo model & LLM Guard. This innovative approach not only enhances the overall security and integrity of the AI system but also ensures ethical and safe user engagement by implementing intelligent guardrails that can detect and prevent potentially harmful or manipulative input strategies in real-time.

Developed a custom Retrieval-Augmented Generation (RAG) framework for search application for handling multimodal embeddings for diverse data structures (PDFs, documents, Excel, CSV, PPTs, images) with tailored chunking, retrieval strategies along with Ranking models using Cross Encoders with Elastic Search. Built tools, agents, and graphs leveraging a multi-agentic architecture along with automating end to end data pipeline using Apache Airflow Dags with schema detection and schema Validation.

Executed end-to-end deployment of a multi-agent application, including containerization with Docker Registry, model serving via Kubeflow (K-Serve) within a Kubernetes cluster, and pipeline automation using Jenkins. Integrated logging and monitoring of metrics across different pods using Prometheus, Loki and Grafana.

Worked on creating a code-generation app with Reinforcement Learning with Human Feedback (RLHF) for performance improvement utilizing PPO. Familiarity with techniques like GRPO for training reasoning models.

Automating resource provisioning on AWS using Infrastructure as Code (IaC) with Terraform scripts and Ansible.

Collaborated on projects for clients including Major League Baseball, Volkswagen, and Wegmans, training models for various clubs and TV networks with multiple channels and internal/external factors. Leveraged Adobe Mix Modeler to generate AI-driven spend plans. Successful results motivated the client to onboard the module and extend the contract for another term.

Participated in Hackathon, product enhancement features for AMM with customer support (MLB, Volkswagen, Wegmans), including tasks such as Exploratory Data Analysis (EDA), adoption planning, go-to-market strategy, and demos showcasing Marketing Mix Modeling (MMM techniques & capabilities.

Implementing the developed Gen AI Maturity Assessment model, by initially conducting POC and interviews to understand pain points and assess their Gen AI maturity based on the model, to offer tailored Adobe services.

Experience working in multi-disciplinary teams along with a team player mindset, characterized by effective communication, collaboration, and feedback skills. ERNST AND YOUNG BANGALORE, INDIA

Senior ML Engineer AI Specialist (Aug2022 – Sep 2023) Demonstrated expertise in both structured and unstructured data handling, crafting various Proof of Concepts (PoCs) and Minimum Viable Products (MVPs) utilizing Generative AI and Machine Learning/Deep Learning algorithms.

Implemented an end-to-end chat application using search relevance, ranking, and retrieval systems, including vector search for scanned Lease Documents using Large Language Model (GPT 3.5) and restful API services by improving retrieval quality with embeddings and vector search for customer Ford Motor Company.

Extracted data for 28 product categories to perform EDA, Finetune transformer DistilBERT model to classify 28 categories and deploying model to Azure cloud using Docker and Azure Kubernetes services for CPG customer Pepsico. Extracting Invoice details from invoices using Azure Cognitive services and finetuning LayoutLM Model and automating the process using Robocorp Automation Tool for consumer segmentation for client Philip Morris International. COGNIZANT BANGALORE, INDIA

Awarded with “Doing the right thing, the right way” award as #Cognizant cheer campaign for leading the team in Cognizant on 07 March 2022. Senior Data Scientist (Feb 2021– Aug 2022)

Led a team of 10 engineers as Team lead/mentor shaping the career path and aligning customer requirements for multiple projects.

Using Machine Learning, NLP and analytical skills for handling humongous data for analyzing and predicting total spend of the company using Marketing Mix Modeling (MMM) with various vendors. Identifying the products of the spend and to help them calculating the profits percentage per year and in deciding the budget for coming year by classification and Prediction of the invoices generated for 101 countries across the globe into predefined categories using Azure Databricks with pySpark for customer British American Tobacco with 70%increase in automation and decrease in reporting time from 40 to 10 days.

Automatic Customer Email’s Classification in Italian Language into multiple predefined categories using fine-tuned mBert Model. Simple response generation using GPT-Neo reducing manual efforts to 80% as part of POC for an e-commerce client.

Increased efficiency (improved accuracy and time gains) whilst classifying user comments into pre-defined multiple Core themes for an international retail store using Bert models, Snorkel, and machine learning algorithms like XGBoost, Decision tree, SVM. SRM INSTITUTE OF SCIENCE & TECHNOLOGY INDIA

Assistant Professor (Jul 2018 – Jan 2021)

Taught graduate-level courses on Machine Learning, Deep Learning, Data Mining, Python, Data structures and AI methodologies, receiving accolades for teaching excellence.

Supervised research projects on Machine learning, NLP and Deep Learning for search applications, recommender system etc.

Organized and led industry workshops on emerging AI technologies, fostering collaboration between academia and industry.

LORD KRISHNA COLLEGE OF ENGINEERING UP, INDIA

Assistant Professor (Jul 2010 – Jan 2017)

Taught graduate-level courses on Artificial Intelligence, Compiler Design, Digital Image Processing, Algorithms, SQL. Presented and published research articles in International Journals and conferences. EDUCATION & CREDENTIALS

Pursuing Doctorate (Ph.D) in Computer Science & Engineering, SRM Institute of Science & Technology, Chennai, India, Research Area: Deep Learning, Machine Learning and Natural Language Processing for multilingual classification & search M.Tech in Computer Science, Lingaya’s University, Haryana, India, 2013, CGPA: 9.07/10 Thesis Description: Implementation of SOFM for Analyzing Trends in Satellite Imagery with subjective areas Artificial Neural Network and Image processing using MATLAB.

B.Tech in Computer Science and Engineering, Uttar Pradesh Technical University, U.P., India, 2009, Percentage: 76.98 % PUBLICATIONS & CERTIFICATIONS

Published Paper in International Conference on Emerging Trends in Expert Applications & Security entitled "Comparing Ensemble Techniques for Bilingual Multiclass Classification of Online Reviews" as part of Lecture Notes in Networks and Systems (LNNS, volume 681), under Springer Nature Singapore, p.p. 275-284, online from 13, June 2023.

Published Paper in International Conference on Intelligent Human Computer Interaction (IHCI) entitled "Multiclass Classification of Online Reviews using NLP & Machine Learning for Non-English Language" as part of Lecture Notes in Computer Science (LNCS, volume 13741), under Springer Nature Switzerland, p.p. 85-94, online from 11, April 2023.

Completed Robocorp Certification for Level 1, 2 and 3 for Automation Developer in January2023.

NPTEL certified in “Machine Learning” with “ELITE” conducted by IIT Kharagpur, India from June 2018 to Nov 2018.

NPTEL certified in “Data Mining” with “ELITE and SILVER MEDAL” conducted by IIT Kharagpur, India from Dec 2018 to May 2019.

Presented Paper in International Journal of Emerging Technology and Advanced Engineering (IJETAE) entitled

"Implementation of Artificial Neural Network (SOFM) for future prediction in Satellite Imagery" in Volume 5, Special issue 1, April 2015.

Presented Paper in International Journal of Emerging Technology and Advanced Engineering (IJETAE) entitled

“Amalgamation of Smartphones using Byte Code Level Cross Compilation” in Volume 4, Special issue 1, February 2014.

Published Paper in International Journal of Soft Computing and Engineering (IJSCE) entitled “Analysis of Satellite Images using Artificial Neural Network” in Volume-2, Issue-5, and January 2013.

Published Paper in International Journal of Engineering and Science Research (IJESR) entitled “Implementation of Self Organizing Feature Map for Analyzing Trends in Satellite Imagery” in Volume-3, Issue-6, and May 2013.

Published Paper in International Conference on Advanced Computing & Communication Technologies organized by Asia pacific Institute, Panipat on the topic “Byte Code Level Cross Compilation for Amalgamation of Smartphone” held on 3 Nov. 2012.

Presented Paper in National Conference held in H.R Institute of Technology on the topic “Voice Over Internet Protocol

(VOIP)” in year 2009.

Contact this candidate