Ajay Therala

AI Engineer, Software Developer & Researcher

Innovative AI Engineer and Software Developer with expertise in Generative AI, scalable data pipelines, and end-to-end ML solutions. Passionate about building intelligent systems, optimizing performance, and solving complex problems at scale. Experienced in research, software development, and deploying AI-driven applications in real-world scenarios.

Education

Master of Science in Computer Science

Arizona State University, Tempe, Arizona, United States.

2023 - 2025

G.P.A - 4.0

Bachelor of Technology in Information Technology

Jawahrlal Nehru Technological University, Hyderabad, Telangana, India.

2017-2021

G.P.A - 9.21

Technical Skills

Programming Languages

Python PySpark SQL C Spark SQL

Web Technologies

Java Script HTML/CSS Django FastAPI Streamlit

Data Science & ML

TensorFlow PyTorch Scikit-learn Pandas NumPy

DevOps & Cloud

AWS Git CI/CD Linux

Databases

PostgreSQL SQL Server MongoDB DynamoDB Elasticsearch OpenSearch

Software Development

Data Structures & Algorithms Object Oriented Design

Publications

Multimodal Large Language Models across Distribution Shifts and Augmentations

NetMD- Network Traffic Analysis and Malware Detection

Projects

Analysis of Yelp Dataset

Optimized Yelp dataset processing by converting JSON to Parquet, leveraging PySpark & Spark SQL for scalable ETL, data transformations, and complex queries in a distributed environment.

Python Data Visualization PySpark Spark SQL

AI Resume Analyzer

Developed an AWS-powered AI Resume Analyzer leveraging Amazon Bedrock's LLM, API Gateway, and Lambda to assess job fit, identify skill gaps, and provide AI-driven career insights through a Streamlit-based UI.

Python AWS StreamLit

Analyzing and Mitigating Hallucinations in Multi modal LLM's

Evaluated multi-modal LLM's like Instruct BLIP on 15,000+ test cases, leveraging Mistral LLM to generate 75,000+ QA pairs and identifying 47% hallucinated responses in count and color-based questions..

Python Mistral LLM Instruct BLIP

Professional Experience

AI Full Stack Developer

Arizona State University - AI Acceleration Team

Aug 2024 - Present

Tempe, AZ

Optimized Data Ingestion Pipeline - Spearheaded a 95% reduction in file chunking times for large datasets (from 1500 s to 4.87s) by integrating an advanced chunking approach, enhancing data processing efficiency.
Containerized & Deployed a Docker image to AWS ECR, integrated it with AWS Lambda for server-less execution. Proposed & Led the implementation of multiprocessing to enhance pipeline efficiency and fully leverage AWS Lambda's vCPUs.
Crafted a script that leverages the Google Drive API to extract and index course content into Open Search, empowering seamless Retrieval-Augmented Generation (RAG) for fast and precise content discovery.
Developed logic for parsing JSON files, achieving a 30% reduction in ingestion time and enhancing RAG efficiency.
Initiated an exploration of Bedrock's Knowledge Bases to assess their efficiency & potential as a replacement for the current pipeline.

Machine Learning Engineer

Tata Consultancy Services

Aug 2021 - Aug 2023

Hyderabad, TS/p>

Engineered core components for a domain-specific search engine (Cognitive Product Support), enhanced search accuracy by 30%.
Crafted Data Lens, a component for training custom NER models, achieving an impressive 80% - 90% accuracy.
Demonstrated expertise in Generative AI, Prompt Engineering by developing advanced GPT-powered bots handling over 5,000+ interactions daily. Delivered impactful client demos, earning high praise from esteemed clientele.
Automated document processing using AWS Textract, developed a POC for key-value extraction from handwritten forms.

Research Project Intern

Tata Consultancy Services

Jan 2021 - Aug 2021

Remote

Improved detection accuracy by 6% on NetML, CICIDS2017, and non-vpn2016 datasets through data refinement techniques, and Bagging & Boosting algorithms.
Achieved a top 5 position in the NetML Network Traffic Analytics Challenge 2020 and presented research findings at ICAIIC 2022 to over 300 peers and industry professionals.