Arya Kinagi
AI Engineer | Ex-CGI (Software Engineer - Machine Learning & NLP) | MS Computer Science'26 | Published Researcher (IEEE, Springer) | Python | AWS | Scalable AI Pipelines | Data Engineering
Contact Me
Building AI Solutions That Are Reliable
I build end-to-end AI software solutions that drive significant business impact, leveraging deep learning and NLP to solve complex problems from model development to containerized deployment. Beyond engineering, I enjoy finding the signal in the noise through data analysis, uncovering actionable insights from complex datasets.
At CGI, I tackled a workflow overwhelmed by 240,000 annual support tickets. By building an NLP model that understood semantic meaning, my system automatically identified root causes, ensuring critical issues were addressed immediately. I bridge the gap between technical data and business strategy, presenting solutions to management and company founders.
Core Expertise
  • Machine Learning & NLP
  • Data Analysis
  • AWS Cloud Architecture
  • Scalable AI Pipelines
  • Data Engineering & ETL
  • Published Research
Technical Skills
Python & SQL
Proficiency in Python for ML/AI development and SQL for data engineering and analysis.
AWS Cloud
Lambda, API Gateway, DynamoDB, S3 for scalable, serverless architectures supporting production workloads.
LLM & NLP
Large Language Models, natural language processing, semantic analysis, and text classification systems.
React.js
Frontend development with React for building dynamic, user-friendly web applications and interfaces.
ETL Pipelines
Designing and implementing robust data pipelines for extraction, transformation, and loading at scale.
Machine Learning
Deep learning, CNNs, transfer learning, clustering algorithms, and model optimization for production.
Experience
Impact in Action
Here are my Industry Experiences where my technical skills translated into tangible business value and innovative solutions.
Intelligent Document Processing System
June 2025 - December 2025
AI Engineer
Scalable AWS Architecture
Designed and deployed API layer using Lambda, API Gateway, DynamoDB, and S3, enabling stateless request handling and horizontal scaling to support 1,000+ insurance claims/day in production.
LLM Integration
Containerized backend with Docker and integrated self-hosted Mistral-7B LLM, refining prompts and enforcing structured output parsing; reduced processing latency from ~6.2s to ~3.2s per document.
Cost-Effective Solution
Delivered a lower-cost alternative to commercial document AI APIs for clients in just 12 weeks, improving extraction consistency while reducing operational costs.
NLP-Based Ticket Classification System
January 2024 - May 2024
Software Engineer – ML & NLP
The Challenge
ITSM workflows overwhelmed by ~240,000 tickets annually written in varied human language, requiring extensive manual categorization and delaying critical issue resolution.
The Solution
Implemented NLP-based classification service normalizing free-text summaries using CountVectorizer. Trained Multinomial Naive Bayes classifier achieving 85.6% accuracy, integrated with frequency-based logic to surface Incident tickets requiring immediate attention.
240K
Annual Tickets
Processed and classified automatically
20K
Hours Saved
Reduced manual categorization time per year
85.6%
Accuracy
Classification model performance
Advanced Ticket Clustering & Root Cause Analysis
In this project, I Implemented similarity-based grouping using TF-IDF vectors, cosine similarity, and Word2Vec embeddings with NLTK-based preprocessing to cluster tickets by suspected root cause. This end-to-end system became the foundation for the team's production automation tool, enabling proactive identification of systemic issues.
Text Preprocessing
NLTK-based normalization and cleaning
Feature Extraction
TF-IDF vectors and Word2Vec embeddings
Similarity Clustering
Cosine similarity-based grouping
Root Cause ID
Automated pattern detection
Next Chapter
Research & Innovation
Beyond industry applications, my work extends into academic research, pushing the boundaries of AI and contributing to the scientific community.
Research: Advanced Clustering for High-Dimensional Data
KLE Technological University
April 2023 - January 2024
Designed a hybrid unsupervised clustering pipeline combining autoencoder-based dimensionality reduction with K-means optimized using Particle Swarm Optimization (PSO) for high-dimensional data. Implemented autoencoders to learn compact latent representations, mitigating the curse of dimensionality while preserving clustering-relevant structure.
Applied the Elbow Method to determine optimal clusters. Evaluated using Silhouette Coefficient and Davies-Bouldin Index, achieving improved cluster separation over raw features. Validated on NIFTY-100 stock market time-series data.
Research: CNN-Based Disease Detection in Agriculture
KLE Technological University
September 2022 - March 2023
01
Dataset Curation
Custom dataset of healthy and diseased red and green chillies with data augmentation for class balance and robustness.
02
Model Selection
Applied transfer learning across CNN architectures, selected MobileNetV2 for depthwise separable convolutions and edge suitability.
03
High Accuracy
Achieved ~97% test accuracy while maintaining compact model footprint suitable for resource-constrained devices.
04
Mobile Deployment
Converted to TensorFlow Lite and deployed as Android application, enabling on-device inference without cloud dependency.
Professional Journey
1
KLE Technological University
Research Assistant – Agricultural AI | September 2022 - March 2023
Developed CNN-based disease detection system for chillies using transfer learning, achieving 97% accuracy and deployed as TensorFlow Lite Android application.
2
CGI
Frontend Web Developer | June 2023 - September 2023
Developed React-based platform for SHRM's event management with modular UI components and state management.
3
KLE Technological University
Research Assistant – Advanced Clustering | April 2023 - January 2024
Designed hybrid unsupervised clustering pipeline combining autoencoders with K-means optimized using PSO for high-dimensional data analysis.
4
CGI
Software Engineer – ML & NLP | January 2024 - May 2024
Built NLP ticket classification system processing 240,000 tickets annually, saving 20,000 hours of manual work.
5
University of Dayton
Digital Projects Assistant | January 2025 - Present
Contributing to digital transformation of archival magazines, managing large data volumes, and collaborating with library teams on digital asset management.
6
Digiteth Partners
AI Engineer | June 2025 - December 2025
Designed scalable AWS infrastructure and integrated self-hosted LLM for insurance claims processing, reducing latency by 48%.
Education & Publications
Academic Background
University of Dayton
Master's degree, Computer Science
August 2024 - May 2026
KLE Technological University
Bachelor's degree, Computer Science
June 2020 - June 2024
Published Research
Let's Connect: [email protected] | LinkedIn
Made with