~80%
manual reporting effort reduced through LLM-powered automation
Open to AI/ML, LLMOps & data engineering roles
Building production AI
where data meets
reliable systems.
I’m Vasantha Lakshmi Eda, a full-stack AI/ML professional with 4+ years of experience delivering LLM, RAG, machine-learning, and data-engineering systems from architecture through production.
93%
AI output reliability across production inference runs
20+
MBR metrics delivered through automated reporting
10M+
unstructured records processed in NLP research workflows
VEAI / ML
— 00
Hello — I’m Vasantha.
I design, deploy, and operate production AI systems across LLMOps, NLP, machine learning, and data engineering.
My work spans the full AI/ML lifecycle: research, model development, evaluation frameworks, scalable pipeline architecture, and production delivery.
I hold an M.S. in Information Technology from Arizona State University with a 4.0 GPA and am an AWS Certified Machine Learning - Specialty professional.
- Focus
- LLMOps, RAG, ML & data engineering
- Languages
- Python, SQL, R, Go, Java, Bash
- Cloud
- AWS, Azure, GCP
- Delivery
- Docker, Kubernetes, CI/CD, MLOps
- Open to
- AI/ML and data engineering roles
- Contact
- Email Vasantha
Professional experience
AI, data, and automation work from research through production delivery.
Amazon
Data Scientist II
Built LLM-powered reporting, inference, and survey-analysis systems that turned fragmented operational data into actionable decisions.
- Architected an LLM-powered insights system, reducing manual reporting effort by ~80%.
- Orchestrated an AWS Step Functions pipeline across validation, transformation, Bedrock inference, evaluation, and delivery.
- Implemented checks that achieved 93% AI output reliability.
- Deployed reporting for 20+ MBR metrics to targeted audiences.
- Amazon Bedrock
- Step Functions
- SageMaker
- Python
- SQL
CMG Financial
Agentic AI Prompt Engineer
Delivered AI-powered search, conversational simulation, and event-driven transcript automation.
- Engineered a RAG mortgage-policy search system across 40+ policy documents.
- Launched a conversational paydown simulator serving 100+ end users.
- Built a transcript automation workflow for priority extraction and summarization.
- LangChain
- Azure OpenAI
- Azure AI Search
- Python
- n8n
GoTackle
Data Engineer
Volunteer role through the Arizona State University Professional Training Program.
- Designed serverless pipelines using AWS Lambda, Redshift, and BigQuery.
- Built AI-powered business-insight engines with OpenAI GPT, Vertex AI, and LangChain.
- Automated transformations for 10,000+ SMB users.
- AWS Lambda
- BigQuery
- Airflow
- dbt
- Vertex AI
Arizona State University
Research Assistant
- Built a knowledge-graph RAG pipeline that improved information retrieval by 35%.
- Developed NLP workflows processing 10M+ unstructured records.
- Developed a data-matching model that increased accuracy by 25%.
- RAG
- Hugging Face
- spaCy
- Transformers
- Python
Cardlytics
Data Science Analyst
- Built a Streamlit application that improved merchant-location processing efficiency 4x.
- Improved forecasting with XGBoost, LSTM, and LightGBM, reducing operational risk by 25%.
- Developed batch and real-time ETL pipelines with Python, PySpark, AWS, and Kafka.
- Streamlit
- PySpark
- Kafka
- XGBoost
- FastAPI
Selected work
Representative systems from production roles, research, and competition work.
P/01
LLM-powered insights automation
An end-to-end reporting system combining validation, processing, Bedrock inference, evaluation, and targeted delivery.
~80%less manual reporting
93%AI reliability
- Bedrock
- Step Functions
- SageMaker
- Python
P/02
Mortgage policy search
A RAG-based policy search experience for querying mortgage guidelines in natural language across more than 40 policy documents.
40+policy documents
- LangChain
- Azure OpenAI
- Azure AI Search
P/03
Knowledge-graph RAG
A semantic-search and contextual-recommendation pipeline using knowledge-graph triplets.
35%retrieval improvement
10M+records handled
- Transformers
- spaCy
- Python
- RAG
P/04
Survey-analysis pipeline
An automated NLP pipeline for survey ingestion, cleaning, sentiment analysis, reason grouping, and CSAT reporting.
7+tools analyzed
~60%less manual effort
- Python
- NLTK
- spaCy
- QuickSight
P/05
AI-driven NLP solution
A real-world NLP solution recognized with a Smart India Hackathon 2021 win.
WinnerSmart India Hackathon
- NLP
- Machine Learning
Technical stack
A systems view of the tools I use to move AI work from idea to dependable delivery.
From signal to system
Built for the full
AI lifecycle.
I connect model development, data infrastructure, cloud services, and delivery practices into production-ready systems.
- 01Data & context
- 02Models & retrieval
- 03Evaluation & orchestration
- 04Delivery & observability
01AI systems
LLMs &
intelligence
Production RAG, evaluation, prompting, and model development.
- Amazon Bedrock
- Azure OpenAI
- LangChain & LangGraph
- Transformers
02Data foundation
Data &
machine learning
- Python, SQL, R
- PySpark & Kafka
- scikit-learn
- XGBoost & LightGBM
03Production layer
Cloud &
MLOps
- AWS, Azure, GCP
- SageMaker Pipelines
- Docker & Kubernetes
- GitHub Actions
04Decision layer
Analytics &
insight
- Tableau & Power BI
- QuickSight
- Statistics & A/B testing
- Model evaluation
Education & credentials
Academic foundation, certification, and community involvement.
Education
M.S.
Information Technology
Arizona State University
GPA: 4.0 / 4.0
B.Tech
Information Technology
Jawaharlal Nehru Technological University
GPA: 8.3 / 10
Certification
- AWSAWS Certified Machine Learning - Specialty
Activities
Hackathon Winner
Smart India Hackathon 2021
Built an AI-driven NLP solution under real-world constraints.
Open Source Contributor
Shared ML notebooks and solutions on Kaggle.
AI Community Member
Engaged with MLOps, coding, and papers through Code.
Let’s build
something useful.
Open to conversations about production AI, LLMOps, data systems, and ML engineering opportunities.