AI Engineer specializing in Agentic AI Development, Large Language Models, and Production-Grade AI Systems. I architect and deploy intelligent agents using Google ADK (Gemini API) that combine LLMs with real-time data processing, tool use, and autonomous decision-making capabilities.

Core Expertise:

• 🤖 Building agentic AI systems with Google ADK, Gemini models, and function calling
• 🔄 Implementing RAG (Retrieval-Augmented Generation) architectures with vector databases
• 🛠️ Developing function calling agents with dynamic tool use and real-time data integration
• ☁️ Deploying scalable AI pipelines on GCP (Vertex AI, Cloud Run, Firebase)
• 🎯 Production ML model serving, monitoring, and optimization on Google Cloud

• Real-time AI Agents: Built production agents using Google ADK with function calling for autonomous task execution
• Function Calling & Tools: Implemented dynamic function calling with Gemini models for real-time data fetching and tool orchestration
• RAG Systems: Architected enterprise-grade RAG pipelines with 45% accuracy improvement using Pinecone vector stores
• Prompt Engineering: Extensive experimentation with Gemini models optimization, few-shot learning, and structured output generation

• Production ML Pipelines: Deployed models on GCP Vertex AI with automated retraining and versioning
• Serverless AI: Containerized ML services on Cloud Run for low-latency inference at scale
• Model Monitoring: Implemented performance tracking, drift detection, and automated alerting systems
• Cost Optimization: Reduced inference costs by 35% through efficient batching and caching strategies

May 2024 - Present | Coventry, UK

Building an agentic AI platform for live video assistance with autonomous decision-making capabilities:

Agentic AI Architecture:

• Architected multi-agent system using Google ADK with function calling for autonomous task execution
• Implemented dynamic function calling with Gemini models for real-time tool selection and data retrieval
• Built agent orchestration layer handling planning, execution, and error handling
• Developed context management systems for maintaining conversational state across sessions

RAG & Vector Search:

• Engineered production RAG pipeline with Pinecone vector database (45% accuracy improvement)
• Implemented semantic chunking strategies and hybrid search (dense + sparse retrieval)
• Built real-time embedding generation using Gemini embedding models and vector upsert pipelines
• Optimized retrieval with metadata filtering and re-ranking algorithms

ML Infrastructure & Deployment:

• Deployed ML models on GCP Vertex AI with automated model versioning and A/B testing
• Built FastAPI microservices with async processing for real-time Gemini API inference
• Implemented Cloud Run auto-scaling for handling variable AI workloads
• Created monitoring dashboards for tracking model performance, latency, and costs

Computer Vision Integration:

• Integrated YOLO object detection for real-time video stream analysis
• Built frame extraction and preprocessing pipelines using OpenCV
• Combined vision outputs with Gemini multimodal capabilities for enhanced reasoning

Key Technologies: Python, Google AI SDK (Gemini API), Pinecone, FastAPI, Vertex AI, Cloud Run, Firebase, Next.js, TypeScript

October 2023 - May 2024 | Rugby, UK

AI/ML Development:

• Built data processing pipelines with Pandas and NumPy for ML feature engineering
• Developed RESTful APIs using FastAPI for serving ML predictions
• Automated workflows using Python scripts, reducing processing time by 60%

Cloud & DevOps:

• Containerized Python applications with Docker and deployed to Azure Cloud
• Implemented CI/CD pipelines for automated testing and deployment
• Integrated external systems using REST APIs and GraphQL

Key Technologies: Python, FastAPI, Django, React, TypeScript, Next.js, Docker, Azure, Pandas, NumPy

February 2021 - August 2022 | Bangalore, India

ML Model Development:

• Designed and deployed ML classification models using Scikit-learn and TensorFlow for incident categorization (40% efficiency gain)
• Built predictive models for anomaly detection using supervised learning techniques
• Performed feature engineering, hyperparameter tuning, and model evaluation

Data Engineering:

• Developed Python automation scripts for data migration to Azure Cloud Storage
• Created Jupyter notebooks for data analysis, visualization, and model experimentation
• Built data pipelines for processing operational metrics

Key Technologies: Python, TensorFlow, Scikit-learn, Azure, Jupyter, Pandas, NumPy

Capabilities:

• Agentic systems with Google AI SDK and function calling
• Multi-turn conversations with context management
• Dynamic tool orchestration and execution
• Structured output generation with Gemini models
• Gemini embedding models for semantic search
• Some experience with LangChain for agent workflows

Experience with:

• Semantic search and similarity matching
• Hybrid search strategies (dense + sparse)
• Metadata filtering and query optimization
• Embedding model selection and optimization

Applications:

• Real-time object detection and tracking
• Video stream processing and analysis
• Image classification and segmentation
• Multimodal AI (vision + language)

GCP Services:

• Vertex AI (model training, deployment, monitoring)
• Cloud Run (serverless container deployment)
• Cloud Functions, Cloud Storage, BigQuery
• Firebase Authentication, Firestore

Technologies: Python | LangChain | OpenAI API | Gemini | Pinecone | FastAPI | OpenCV | YOLO | Vertex AI | Cloud Run

Agentic AI Architecture:

• Built autonomous agent system using LangChain with tool-calling capabilities for real-time decision making
• Implemented Model Context Protocol (MCP) for dynamic data fetching and processing
• Developed agent memory systems for maintaining context across multi-turn conversations
• Created planning and reasoning loops with self-correction mechanisms

Production RAG System:

• Architected end-to-end RAG pipeline with Pinecone vector database
• Implemented semantic chunking and embedding strategies for optimal retrieval
• Built hybrid search combining dense vector similarity and sparse keyword matching
• Result: 45% improvement in answer accuracy

ML Deployment:

• Deployed on GCP using Vertex AI for model serving and Cloud Run for microservices
• Implemented real-time video processing with OpenCV and YOLO object detection
• Built FastAPI backend with async processing for low-latency inference
• Created monitoring dashboards for tracking model performance and costs

Impact: Production-ready agentic AI system handling real-time video analysis and intelligent assistance

Technologies: Python | TensorFlow | OpenCV | YOLO | NumPy

Computer Vision ML:

• Developed YOLO-based object detection model for real-time traffic violation detection
• Implemented end-to-end ML pipeline: data collection → annotation → training → deployment
• Built video processing pipeline using OpenCV for frame extraction and analysis
• Performed model optimization and quantization for edge deployment

Results:

• 87% detection accuracy on test dataset
• Real-time processing at 30 FPS
• Successfully detected helmet violations and traffic rule breaches

Publication: IEEE - "An Intelligent Traffic Monitoring System for Non-Helmet Wearing Motorcyclists Detection"

Technologies: Python | Scikit-learn | Pandas | NumPy | Jupyter

Machine Learning Development:

• Built classification model for predicting B-cell epitopes of alphavirus
• Performed extensive feature engineering and data preprocessing
• Conducted hyperparameter tuning and cross-validation
• Implemented ensemble methods for improved predictions

Result: 15% improvement in prediction accuracy over baseline models

Technologies: Django | PyTorch | Transformers | REST API

Deep Learning NLP:

• Developed transformer-based model using PyTorch for document understanding
• Built NLP pipeline for extracting entities and relationships from requirement documents
• Created RESTful API using Django REST Framework for model inference
• Automated UML diagram generation from natural language specifications

Application: Automated software design documentation generation

Technologies: Python | TensorFlow | Scikit-learn | Azure

Enterprise ML Deployment:

• Designed and deployed classification model for automated incident categorization
• Built predictive models for system anomaly detection
• Implemented model monitoring and retraining pipelines
• Integrated with enterprise systems via REST APIs

Impact: 40% reduction in manual categorization effort

MSc Artificial Intelligence
Aston University, Birmingham, United Kingdom | September 2023

B.Tech Computer Science and Engineering
APJ Abdul Kalam Technological University, Kerala, India | February 2021

"An Intelligent Traffic Monitoring System for Non-Helmet Wearing Motorcyclists Detection"
IEEE Conference | View Publication

AI/ML Specialized:

• 🎓 Artificial Intelligence Masters - Simplilearn
• 🎓 Deep Learning with TensorFlow - IBM
• 🎓 Fundamentals of Deep Learning for Computer Vision - Nvidia Deep Learning Institute

Cloud & Infrastructure:

• ☁️ Google Cloud Fundamentals: Core Infrastructure - Google
• ☁️ Reliable Google Cloud Infrastructure: Design and Process - Google
• ☁️ Azure Fundamentals - Microsoft

Programming:

• 💻 C# Certification - TestDome
• 🐍 Python Training - IIT Bombay

I'm seeking AI Engineer positions focused on:

• 🤖 Agentic AI development with LangChain, LangGraph, and autonomous agent systems
• 🔄 RAG architecture design and optimization for production environments
• 🚀 LLM deployment and MLOps on cloud platforms (GCP, Azure, AWS)
• 🛠️ Building production-grade AI systems with real-world impact
• 🧠 Multimodal AI combining language, vision, and structured data

My Ideal Role: Building intelligent, autonomous AI systems that solve complex problems through agentic reasoning, tool use, and seamless integration with existing infrastructure.

I'm always interested in discussing AI engineering, agentic systems, and production ML deployment!

• 📧 Email: aswinkrishna979@gmail.com
• 💼 LinkedIn: linkedin.com/in/ashwin-mariyathu-krishnakumar
• 🐙 GitHub: github.com/ashwinkrishna979
• 📱 Phone: +44 7833846064