Hi there, I'm Gokul Krishna 👋 Machine Learning Engineer | MLOps Specialist | AI Workflow Architect 🚀 As a passionate ML engineer, I specialize in building scalable, production-grade AI solutions with a focus on automation, reliability, and performance. I thrive at the intersection of data science and DevOps, where engineering meets intelligence. 🧠 With hands-on experience in predictive modeling, real-time inference APIs, and automated training pipelines, I enjoy solving real-world problems using tools like FastAPI, MLflow, DVC, Celery, and AWS. 🎯 Whether it’s deploying intelligent systems, optimizing simulation-based models, or orchestrating end-to-end pipelines — I aim to build AI that not only works, but works smart. 📌 Explore some of my featured projects below to see my work in action.

A serverless, AI-powered blog generation system featuring:

• Amazon Bedrock foundation models for AI text generation
• AWS Lambda for serverless orchestration
• Amazon S3 for persistent blog storage
• API Gateway for secure, HTTPS-based client access
• Fine-grained IAM roles for secure Bedrock and S3 integration

🔗 Tech: AWS Lambda, Amazon Bedrock, API Gateway, Amazon S3, Python
📂 Design: Serverless, cloud-native, scalable, production-ready

A production-grade, end-to-end abstractive text summarization system featuring:

• Full fine-tuning of Google Pegasus LLM for domain-specific summarization.
• Modular, MLOps-ready pipelines for ingestion, validation, transformation, training, evaluation, and prediction.
• FastAPI backend with Jinja2-powered web UI.
• DVC dataset versioning, MLflow experiment tracking, and centralized logging.
• Local + AWS S3 artifact storage with YAML-driven configuration.

🔗 Tech: FastAPI, Hugging Face Transformers, Google Pegasus, DVC, MLflow, AWS S3, YAML
📂 Design: Modular, reproducible, cloud-ready, production-grade

A fully modular, production-grade ML pipeline for phishing detection with:

• FastAPI for API serving (/train, /predict)
• Async training using Celery + Redis
• Workflow versioning with DVC
• Hyperparameter tuning via Optuna
• Experiment tracking using MLflow + DagsHub
• Schema + drift validation, preprocessing, and artifact management
• CI/CD with GitHub Actions, auto-sync to AWS S3

🔗 Tech: FastAPI, Scikit-learn, Celery, MLflow, Optuna, AWS, MongoDB, Docker
📂 Design: Modular, YAML-driven, CI/CD-enabled

A reusable and scalable machine learning pipeline to predict student scores:

• PostgreSQL-based ingestion with table auto-creation from YAML schema
• Configurable preprocessing (scaling, encoding, imputing, column operations)
• Full pipeline: ingestion ➜ validation ➜ transformation ➜ training ➜ evaluation ➜ prediction
• Optuna-powered hyperparameter tuning with MLflow tracking
• Centralized logging with AWS S3 and local backups
• Artifact versioning using DVC

🔗 Tech: Scikit-learn, Optuna, PostgreSQL, MLflow, DVC, AWS, YAML
📂 Design: Production-grade, modular, reusable, data-driven, CI/CD-enabled

A Dockerized regression pipeline with a simple web interface for real-time prediction:

• Built using ElasticNet Regression
• Flask frontend for user input and result display
• Modular pipeline: ingestion → validation → transformation → training → prediction
• Structured with reusable entity-based config classes
• MLflow-based experiment logging & reproducible config

🔗 Tech: Flask, Scikit-learn, MLflow, Docker, YAML
📂 Design: Lightweight, extensible, production-ready

🔹 Data Scientist / CAE Engineer – General Motors (via TCS)
Sep 2019 – Aug 2023 · Bangalore, India

• Reduced EV battery module mass by 11% via ML-driven simulation optimization
• Automated FEA workflows and built predictive models for unseen load conditions
• Developed Kriging-based optimization and mentored junior engineers in ML-integrated engineering

🔹 Structural Analyst – Johnson & Johnson MedTech (via TCS)
Feb 2017 – Aug 2019 · Kolkata, India

• Optimized design cycles for surgical devices using FE analysis and statistical tools
• Built automation scripts for structural simulation and data extraction
• Supported design for novel uterine tumor excision device using biomechanical modeling

• 🛠 I’m currently working on
  Production-grade AI APIs with FastAPI + Celery + MLflow for scalable ML deployment.

• 🤝 I’m looking to collaborate on
  Applied ML projects, MLOps tooling, or LLM use cases in healthcare or finance.

• 🧠 I’m looking for help with
  Model serving at scale and efficient Kubernetes-based deployment.

• 🌱 I’m currently learning
  LLM fine-tuning, Kubeflow pipelines, and advanced ML monitoring strategies.

• 💬 Ask me about
  MLOps pipelines, Optimization, Automation, or real-world AI deployment.

• ⚡ Fun fact
  My first ML pipeline was trained entirely on FEA simulation data—no labeled dataset, just raw physics!

: nv-gokul-krishna

: nvgokulkrishna@gmail.com