01
Retrieval (RAG)
I build retrieval pipelines over internal documents with pgvector and FAISS, and I choose embedding models and chunking strategies to improve search quality. The goal is more relevant retrieval and fewer made-up answers.
Backend engineer, Dallas area
I build AI agents and retrieval systems for enterprise teams.
Backend engineer working with Python, RAG, MCP and the systems around them: APIs, databases and access control.
Ask about my work
Answers come only from my resume and project text, with sources shown.
Try one:
0 of 500
How the chat works
Rate limit
Limits how often one visitor, and the site as a whole, can ask questions.
Upstash RedisSource for Rate limit: lib/ratelimit.tsInput check
Refuses hidden or lookalike characters, and questions that try to script the answer, before any search or model call.
Plain TypeScriptSource for Input check: lib/chat/answer.tsRetrieval
Turns the question into an embedding and finds the closest passages of my resume.
pgvector in NeonSource for Retrieval: lib/retrieve.tsScope check
Refuses the question if even the best passage is not close enough to be about my work.
Similarity thresholdSource for Scope check: lib/chat/answer.tsModel
Writes an answer from the retrieved passages only, and names the passages it used.
Answer checks
The server confirms every citation was retrieved, then blocks unlisted links, private words, and numbers or names the sources do not state.
Server-side checksSource for Answer checks: lib/chat/answer.ts
Every refusal returns the same fixed message.
Focus
02
Agents and tool use
I build agents that plan multi-step work and call tools across enterprise systems. I also build the React dashboards people use to monitor and check what the agent did.
03
Secure integrations
I connect language models to internal data through MCP servers and REST APIs. I add authentication, rate limiting and access controls to the servers and endpoints that handle sensitive data.
Experience
February 2025 to present
Software Development Engineer
American Airlines. Dallas, TX (remote).
Backend-focused full stack work on AI tools for operations, customer support and internal automation.
Details
AI systems
- Built Python services with FastAPI and integrated language models (OpenAI, Azure OpenAI) into internal tools to automate business workflows.
- Built RAG pipelines over internal knowledge bases with pgvector and FAISS, and evaluated embedding models and chunking strategies to improve search.
- Built agents for multi-step workflows: task planning, tool calling and orchestration.
- Used ML techniques (feature engineering, model evaluation) in systems that combine traditional ML models with LLM-based reasoning.
- Used few-shot prompting, chain-of-thought structuring and system prompt design to make answers more accurate and consistent.
Integrations and security
- Built MCP servers to standardize secure access to internal data sources, so several AI tools could reuse the same context.
- Added authentication, rate limiting and access controls to MCP servers and REST API endpoints that handle sensitive data.
- Connected AI services to core enterprise systems through REST APIs.
Data and operations
- Used PostgreSQL and MongoDB for application context, conversation state and retrieval, and tuned query performance.
- Monitored AI applications in production to find latency and accuracy problems in responses and agent workflows.
- Built React interfaces and dashboards so business users could monitor and check agent output.
Team and standards
- Set internal standards for MCP integration patterns and prompt versioning.
- Wrote guides for engineers new to RAG and agent development.
- Presented designs and trade-offs to engineering leadership and product.
- Worked with data science, platform and product teams in Agile/Scrum, with Git and CI/CD.
January 2024 to December 2025
M.S., Applied Statistics and Data Science
University of Texas at Arlington.
Full-time study from January 2024. Joined American Airlines in February 2025 while finishing the degree. CGPA 3.9.
June 2021 to December 2023
Software Engineer
American Express. Hyderabad, India.
Backend services and APIs for fraud detection, risk scoring and customer analytics.
Details
Services and data
- Built backend services and REST APIs with Flask and FastAPI.
- Modeled and maintained SQL (PostgreSQL, MySQL) and MongoDB data structures.
- Built full stack features with React and Python APIs for internal risk and compliance tools.
Machine learning in production
- Built data pipelines that brought scikit-learn models into backend services for real-time fraud scoring and risk classification.
- Worked with data science teams to move models from prototype into production.
- Supported early NLP tooling that classified customer support tickets.
Engineering practice
- Refactored legacy Python services into modular, testable components.
- Automated recurring data validation and reporting with Python scripts.
- Debugged production incidents with on-call and platform teams.
- Documented API contracts with OpenAPI/Swagger and reviewed pull requests.
- Followed code review, unit and integration testing, and CI/CD with Git and Jenkins.
Skills
Show skills by area
Backend
- Python
- FastAPI
- Flask
- Django REST Framework
- REST API design and integration
- Microservices
- asyncio
AI and LLM engineering
- LLM integration
- RAG
- AI agents
- Agent orchestration
- MCP
- Prompt engineering
- LangChain
- LlamaIndex
Machine learning
- scikit-learn
- PyTorch (working knowledge)
- Pandas
- NumPy
- Model evaluation
- Feature engineering
- Statistical modeling
LLM platforms and APIs
- OpenAI API
- Azure OpenAI Service
- Anthropic Claude API
- AWS Bedrock
Vectors and search
- pgvector
- FAISS
- Pinecone
- Chroma
- Embedding models
- Semantic search
Frontend
- React
- Redux
- TypeScript
- JavaScript
- HTML
- CSS
Databases
- PostgreSQL
- MySQL
- MongoDB
- DynamoDB
- Redis
- SQL
Cloud and DevOps
- AWS (Lambda, EC2, S3, Bedrock)
- Azure
- Docker
- Kubernetes
- CI/CD
- Jenkins
- GitHub Actions
Practices
- Git
- GitHub
- Agile/Scrum
- OpenAPI/Swagger
- Postman
- Unit and integration testing
- Code review
Projects
Enterprise Knowledge Assistant
One place for employees to ask about internal documentation, policies and system data: retrieval with pgvector, an agent layer for follow-up actions, and a custom MCP server.
AI tools for enterprise operations
Built Python services and AI features for operations, customer support and internal automation; RAG pipelines over internal knowledge bases; agents that automate multi-step workflows; MCP servers for secure access to internal data; React dashboards to monitor and check agent output. Also set internal standards for MCP integration patterns and prompt versioning.
Fraud and risk backend services
Built backend services and APIs for fraud detection, risk scoring and analytics; connected machine learning models to real-time scoring; worked with data science teams to move models from prototype to production.
About
I am a backend engineer with 4+ years of experience, working mostly in Python, and most of my recent work is AI engineering. At American Airlines I own the architecture and delivery of internal AI tooling: RAG pipelines, AI agents and MCP integrations. Before that I worked at American Express on fraud detection and risk scoring services, where I learned what production means with high transaction volumes and real-time scoring.
I care most about problems where getting it wrong has a real cost. Access control that is enforced at the data layer, not only in a prompt. Retrieval that holds up as the document index grows. Evaluation that tells you something is broken before users do.
This site is a small example of the same habits. The chat answers only from my resume and project text. It reads the database through a read-only role, and the server checks that every source it cites is real. I wrote the evals before the chat.