👋 Hi, I'm

Sai Sakalam

Data Engineer @ USAA × IEEE Author × Open-Source Maintainer

10+ years building production data pipelines — from enterprise ETL and Hadoop to Snowflake and dbt in the cloud.
I publish research on data governance and ship open-source tools for Python developers.

See Recent Work LinkedIn ↗

10+ Years in Data

IEEE Journal Paper

PyPI Open-Source Package

scroll

/ recent wins · 2026

📄

Published · IEEE Access T-RKG: regulatory conflict detection with knowledge graphs Peer-reviewed · open access · interactive explorer

↘ 🩹

Released · PyPI pyfirstaid — my first open-source Python package 12 checks · Windows, macOS, Linux · Python 3.9–3.14

↘

/ research & publications

Published Research

Working in regulated data environments taught me that governance problems are real engineering problems. This research takes one of them — conflicting record-keeping rules across systems — and makes it detectable.

IEEE Access 2026 🔓 Open Access

T-RKG: A Knowledge-Based System for Cross-System Regulatory Conflict Detection in Enterprise Records Governance

Kishore Pulagam, Sai Gavaskar Sakalam

IEEE Access, vol. 14, pp. 138420–138439, 2026 · DOI 10.1109/ACCESS.2026.3729794

Large enterprises keep records across many systems, each governed by its own retention and compliance rules — and those rules can quietly contradict each other. T-RKG models regulations, records and systems as a knowledge graph so these cross-system conflicts can be found automatically instead of during an audit.

🧪 Fully open: the ontology, experiment scripts, tests and results are on GitHub so anyone can reproduce or extend the work.

Knowledge GraphsData GovernanceRecords ManagementRegulatory Compliance

Try the interactive Explorer ↗ Read on IEEE Xplore ↗ Code, ontology & experiments ↗ My fork ↗

@article{pulagam2026trkg,
  author  = {Pulagam, Kishore and Sakalam, Sai},
  title   = {{T-RKG}: A Knowledge-Based System for Cross-System Regulatory Conflict Detection in Enterprise Records Governance},
  journal = {IEEE Access},
  volume  = {14},
  pages   = {138420-138439},
  year    = {2026},
  doi     = {10.1109/ACCESS.2026.3729794}
}

See the idea in 5 seconds

Illustrative example: one customer, two systems, two regulations.

0 conflictsEach system checks only its own rules, so everything looks fine.

In the repo's composition experiment, splitting the graph per system dropped detected cross-system conflicts to zero on every run. That gap is what T-RKG closes.

🔍 Want to go deeper? Open the T-RKG Explorer Edit a real sample record, switch siloed vs composed, filter conflicts by severity, and watch a legal hold spread across related emails. ↗

/ open source

pyfirstaid 🩹 first aid for broken Python

"It works on my machine" usually means a broken environment — pip installing into a different Python, SSL errors behind a corporate proxy, a file that shadows the standard library. pyfirstaid runs one command and tells you what is broken and exactly how to fix it.

~/project

$ python -m pyfirstaid
pyfirstaid: checking your Python environment
✔ Python 3.13.1 (supported until October 2029)
✔ Virtual environment active: ~/project/.venv
✘ `pip` installs into a DIFFERENT Python
    fix: use `python -m pip install <package>`
! `random.py` in this folder hides the stdlib module
    fix: rename it, e.g. my_random.py
✔ HTTPS to pypi.org works
✔ 12 compiled package(s) match this Python
✔ No dependency conflicts (pip check)

1 problem(s), 1 warning(s).

—Latest version

—GitHub stars

—Last release

Live from PyPI and GitHub

pip install pyfirstaid

12 checks — pip/python mismatch, venv, PATH leaks, SSL, compiled extensions, broken installs, encoding and more
Cross-platform — tested in CI on Windows, macOS and Linux, Python 3.9–3.14
Works even when pip is broken — a single-file .pyz download runs anywhere
Born from the community — started after a discuss.python.org thread advised building it as a package first

GitHub ↗ PyPI ↗ Report a problem you hit ↗

/ about me

Data engineer by trade,
builder & researcher by nature.

I've spent 10+ years in data engineering — starting with enterprise ETL tools like IBM DataStage and Informatica, working through large Hadoop clusters and on-prem databases, and eventually moving into the modern cloud era with Snowflake and dbt.

Years in regulated environments got me interested in data governance — the rules behind the data. That curiosity turned into a peer-reviewed paper in IEEE Access. Separately, the Python environment problems I kept fixing for colleagues turned into pyfirstaid, my first open-source package on PyPI.

Today I'm a Data Engineer at USAA. I also go deep on how LLMs work and build proofs-of-concept for where AI agents genuinely help in data work — and where they don't.

"Reliable pipelines, clear governance, and tools people actually use."

🩹

Open-Source Maintainer

Created and maintain pyfirstaid — one command to diagnose and fix broken Python environments.

🤖

AI POC Builder

Builds LLM-agent proofs-of-concept for data tasks — the go-to person on my team for AI exploration.

⚡

Data Engineer @ USAA

10+ years shipping production pipelines across on-prem and cloud — DataStage, Hadoop, Snowflake and dbt.

📄

Published Author

Co-author of T-RKG in IEEE Access (2026) — knowledge graphs for regulatory conflict detection.

/ core skills

What I Bring

❄️

Snowflake & dbt

Writing and maintaining dbt models, tests, and docs
Working with Snowflake as the core data warehouse
SQL transformations and data quality checks
Migrating on-prem pipelines to the cloud

Snowflake · dbt · SQL

🤖

AI for Data Engineering

Built a dbt pipeline agent using LLMs
LangChain and function calling for data tasks
Using GitHub Copilot and Claude day-to-day
Team POCs on where AI fits in data workflows

LangChain · Claude · GitHub Copilot

🏢

Enterprise ETL

IBM DataStage — pipeline design and maintenance
Informatica — ETL workflows and mappings
DB2 and Netezza — on-prem data platforms
Hadoop — HDFS, Hive, large-scale batch processing

DataStage · Informatica · Hadoop · DB2

🐍

Python & SQL

Python for data processing and scripting
Advanced SQL — window functions, CTEs, optimization
Pandas for data manipulation and analysis
Writing clean, maintainable transformation logic

Python · SQL · Pandas

Expanding into

⚡ Kafka & streaming 🔥 Spark on Databricks ☁️ AWS data services 🏛️ Lakehouse · Delta & Iceberg 🤖 LLM agents for data 🔬 LLM internals

/ my journey

The Road So Far

2013

Bachelor's in Mathematics

A grounding in analytical thinking that still shapes how I model data and reason about edge cases.

Mathematics

2013 – 2016

ETL Developer

Built ETL pipelines with IBM DataStage and Informatica — learning what production-grade data engineering really demands.

IBM DataStageInformaticaSQL

2016 – 2018

Master's in Computer Science — Lamar University, TX

Moved to the US to study Computer Science at Lamar University, Beaumont. Broadened from data tooling into software engineering fundamentals.

Computer ScienceLamar University

2018 – 2022

Data Engineer — On-Prem Enterprise

Ran large-scale ETL on Hadoop clusters, DB2 and Netezza — real production pressure and legacy complexity.

HadoopDB2Netezza

2022 – 2023

Cloud & Modern Stack

Moved pipelines from on-prem to Snowflake and dbt — still my core stack today.

SnowflakedbtSQL

2023 – 2024

LLMs & AI Agents

Went deep on how large language models work and built first LLM-powered data experiments.

LLMsLangChainPython

2024 – 2025

AI POCs for Data Engineering

Built a dbt pipeline agent and led AI proof-of-concepts for data workflows on my team.

LangChainClaudedbt

2026 · Research 📄

Published in IEEE Access

T-RKG, co-authored with Kishore Pulagam — a knowledge-based system for detecting regulatory conflicts across enterprise record systems. Peer-reviewed, open access.

Knowledge GraphsData GovernanceIEEE

Oct 2026 · Now 🚀