Open to Senior / Staff AI Engineering Roles

Hi, I'm Kasi Vandanapu

Building |

Senior / Staff AI Engineer · Python · LLM Systems · Data & AI Infrastructure

Senior AI & Backend Engineer with 5+ years building production data and AI systems in Python. I work across LLM applications, agentic workflows, data engineering and high-performance backend services — with a focus on making systems faster, cheaper, and easier to evaluate.

Let's ConnectResume
Scroll

About Me

I make data and AI systems faster, cheaper, and measurable.

At Citi, I build LangGraph-based agentic workflows and FastAPI services for AI-powered data applications, including a system that generates dashboard configurations from available data and visualization context.

Separately, I built an open-source Text2SQL evaluation framework that benchmarks LLM-generated SQL against real-world schemas. It analyzes execution and structural correctness, identifies why queries fail, and supports techniques such as schema pruning that reduced inference cost by 50% in my experiments.

Before that, at Infosys, I built ETL pipelines using PySpark, Delta Lake and Airflow for Levi's, Kraft Heinz and HCSC, along with a GCP streaming pipeline using Pub/Sub and Cloud Run. More recently I've been exploring interactive analytics using DuckDB, Parquet and Mosaic — the work documented in my Writing section.

Brandon, FL 33510
5+ Years Experience

Career

Where I've Made Impact

Current Role

Lead Python Developer

Virtusa · Consulting for Citi Bank

Jun 2024 – Present Tampa, FL (Remote)
LangGraphLangChainFastAPILLMsMongoDBAutoGluon
  • Developed LLM-powered agentic workflows within Maphub using LangGraph, LangChain, and LLMs to dynamically generate data models, configure UI elements, and enable conversational data interactions.
  • Designed and implemented microservices with REST APIs using FastAPI to extract, validate, transform, and load data into MongoDB and Oracle databases, supporting inputs from XLSX file and database sources.
  • Built Text2SQL benchmark framework to evaluate and optimize LLM performance on complex enterprise schemas using EX (Execution Accuracy) and QAS (Query Affinity Score) metrics.
  • Optimized schema dump token consumption via intelligent pruning, compression, and selective retrieval — reducing context length and inference costs by 50% while maintaining query accuracy.
  • Integrated real-time ML training and inference using AWS AutoGluon, enabling a fully UI-driven workflow for training, versioning, and deploying live prediction models in Maphub Data Studio.
  • Migrated data validation scripts to Python with vectorized operations, reducing execution time by 60%, deployed as a scalable microservice.

Specialist Programmer

Infosys

May 2019 – Dec 2021 Hyderabad, India
PySparkGCPPub/SubDelta LakeAirflowDatabricks

Work

Featured Projects

Click any card for a full technical deep-dive.

AI/LLMOpen Source
2024–2025

Text2SQL Benchmark

Open-Source Framework

Open-source evaluation framework for testing LLM performance on complex enterprise SQL schemas. Features composite scoring with EX + QAS metrics and intent-aware column selection.

50% token reductionEX + QAS scoringEnterprise schemas
Python ·LLMs ·SQLite +2
Deep dive
AI/LLMHackathon
2024

Wander Finds

Gemini AI Hackathon 2024

AI-driven travel app that generates hyper-personalized travel recommendations using Gemini, Google Maps, and location-aware context.

Gemini multimodalLocation-aware AIHackathon winner
Flutter ·Firebase ·LangChain +2
Deep dive
Data VizAnalytics
2023

Shark Tank Analysis

Interactive Data Dashboard

Interactive D3.js + Tableau dashboard analyzing entrepreneur participation, valuation trends, and deal patterns across all Shark Tank seasons.

All 15 seasonsGeographic heatmapsD3.js force graphs
D3.js ·JavaScript ·Tableau +2
Deep dive
MLNLP
2022

Hate Speech Detection

In-Browser ML Classifier

Real-time hate speech and offensive language detector running entirely in the browser using Pyodide — no server required.

Zero backendWASM inferenceReal-time classification
Python ·XGBoost ·TF-IDF +3
Deep dive

Expertise

Tech Stack & Skill Map

Full-stack AI engineering — from raw data to intelligent, production-deployed systems.

Core Languages

PythonJavaScriptJavaBashSQLC

AI / LLM Stack

LangChainLangGraphLLMsPrompt EngineeringText2SQLAWS AutoGluonXGBoostTF-IDFPyodide

APIs & Microservices

FastAPIREST APIsMicroservicesXLSX ProcessingData Validation

Data Engineering

Apache PySparkDelta LakeApache AirflowDatabricksETL PipelinesStreamingPandasApache ParquetSnapshot Pipelines

Interactive Data Platforms

DuckDB-WASMApache ArrowMosaicObservable PlotColumnar StoragePre-aggregation / OLAP CubesCrossfilteringPredicate PushdownQuery ProfilingWebAssembly

Cloud Platforms

AWSGCPPub/SubCloud RunBigQueryCloud BuildKubernetes

Databases

MongoDBOracleMySQLPostgreSQLDuckDBSQLite

Visualization

TableauD3.jsVega-LiteInteractive Dashboards

Background

Education & Publications

Degrees

Master's in Computer Science

Northeastern University

Jan 2022 – Dec 2023 San Francisco, CA
  • Teaching Assistant — Database Management Systems (Fall '23)
  • Teaching Assistant — Human-Computer Interaction (Summer '23)
  • Research Assistant — List Curator (Summer '22 – Fall '23)

Bachelor of Technology in Computer Science

Amrita School of Engineering

Aug 2015 – May 2019 Coimbatore, India

Publication

“Hadoop and Natural Language Processing Based Analysis on Kisan Call Center (KCC) Data”

2018 International Conference on Advances in Computing, Communication, and Informatics

Vandanapu, K.·2018

Certifications

PCAP™ — Certified Associate Python Programmer

Python Institute

Smart Analytics, ML, and AI on GCP

Google Cloud

Get In Touch

Let's Build Together

I'm open to Senior / Staff Python & AI Engineering roles. Whether you're building an agentic system or scaling a data platform — let's talk.

Email

kasivisu3109@gmail.com

LinkedIn

kasivandanapu

GitHub

kasivisu4

Location

Brandon, FL 33510