Hello! I'm Mayank, a Machine Learning Engineer . My background includes a Masters degree in Machine Learning, with a focus on Deep Learning and Natural Language Processing. I am a Software Engineer with 3 years of work experience in building microservices. I'm skilled at fixing scaling issues, implementing new features, and upgrading microservices to handle more traffic. I'm looking for opportunities to apply my knowledge in Generative AI and Deep Learning to tackle challenging problems.

Feel free to connect with me on LinkedIn or check out my portfolio.

Explore my projects to see detailed applications of machine learning and deep learning across various domains:

A collection of projects demonstrating the use of deep learning techniques in various scenarios:

• Voice-enabled RAG chatbot: Voice-enabled Retrieval-Augmented Generation (RAG) chatbot built on the LLaMA-2 model, integrated with Pinecone for vector search and LangChain for managing the chat capabilities. It is designed to understand and respond to voice queries by referencing content within PDF documents and delivering responses in a synthesized voice.
• Lexical Complexity Predictor: Predicting the complexity of words using transformer-based models like BERT and RoBERTa.
• LLM for Sentiment Analysis of market conditions based on global news reports : Using sentiment analysis to predict market conditions from global news reports.
• Sentiment Analysis of IMDB Reviews: Analyzing sentiments in movie reviews using a Word2Vec model in TensorFlow.
• Semantic Network from Tweets: Building semantic networks from tweets containing specific keywords.
• N-Gram and RNN Text Generator: Text generation using N-Gram models and RNNs with TensorFlow.

A collections of projects demonstrating the use of Backend frameworks.

• SocialMediaApp: A modern, responsive social media platform inspired by Meta's Threads. Built with TypeScript, Next.js, and Tailwind CSS.
• GokuDB: A high-performance database management system written in Go, leveraging powerful concurrency model to achieve efficient multithreaded operations, capable of large-scale data processing . Implemented JSON-based data storage and retrieval, with query handling for rapid access and manipulation of diverse data-sets

• Languages and Libraries: Python, C++, JavaScript, Java, Go, typescript, Sk-learn, Keras, PyTorch, React.js,Next.js
• Database and Frameworks: MongoDB, SQL, DynamoDB, REST API, gRPC, GraphQL
• Machine Learning: RAG , LLM , LoRA, Supervised Fine tuning, Reinforcement Learning , Langchain, Pinecone
• Others: Scikit-Learn, Applied Mathematics, AI, Data Modeling, System Design, Exploratory Data Analysis, Algorithm Development,A/B testing, tensorflow, AWS, Apache Kafka ,PostgreSQL, linux, Data Ingestion, git, kubernetes, Distributed Systems , Jenkins ,AutoML

• Software Development Engineer-2, Scimplify: Engineered scalable, fault-tolerant microservices in fastAPI for internal tools, addressing the product team's requirements for supply chain problems and supporting over 200 manufacturing units.Leveraged Elasticsearch to index products and manufacturers, leveraging edge n-gram tokenization to enable fast, fuzzy search capabilities, reducing p99 latency by 52%. Built a RAG based chatbot by fine-tuning Mistral-7B for an in-house platform using Langchain framework and Pinecone vectorDB to handle customer queries on specialty chemicals and their properties. Deployed on AWS EC2

• Deep Learning Intern, Micron Technology: Spearheaded optimization efforts that led to a 30% improvement in data retrieval speed for constructing abstract syntax trees and control and data flow graphs, enhancing Graph Convolutional Network data feed. Decreased the training time of the model using a multi-GPU cluster for training on distributed systems using PyTorch DDP to reduce training time by 23.6%. Redesigned and improved Deep learning models using PyTorch framework to improve the branch coverage and line-coverage for EDA (Electronic Design Automation) by 17.21% to expedite the semiconductor design process

• BYJU'S: Developed and deployed machine learning models in Python to forecast inventory demand and optimize supply chain logistics, improving the accuracy of inventory management by 15% and reducing delivery times.Developed microservices using Node.js as per REST architecture to meet the requirements of over 2 million inventory requests used for global delivery of retail products. Reduced the latency by 42% using an event driven architecture using Apache Kafka.

• KloudOne: Built and deployed microservices in Go using gRPC for Accuknox, a container security platform.Improved CI/CD (Continuous Integration/Continuous Deployment ) pipeline using Jenkins to reduce downtime by 27%.

• Master of Science in Machine Learning, Stevens Institute of Technology
• Bachelor of Technology in Computer Science and Engineering, Manipal Institute of Technology

• Deep Neural Network Approach for Navigation of Autonomous Vehicles
  • Designed a CNN-based Deep Learning model for navigation of autonomous vehicles using over 200,000 real-world images and sensor data, integrating image compression and normalization techniques for GPU memory optimization
  • Published in IEEE, 6th International Conference for Convergence in Technology as first author. Citations – 5, h- index -1