Resume

Currently @ Schneider Electric · Nashville, TN

hey, i'm steven.

I build data pipelines, AI-powered tools, full-stack apps and systems in C and C++.

Data Engineer / Specialist at Schneider Electric and Computer Science student at MTSU. I turn messy data into reliable pipelines, models and products, from Databricks lakehouses to GradPath, a degree planner that covers all 223 MTSU majors and concentrations.

scroll

~/about me

I studied Computer Science in the Honors program at Middle Tennessee State University, with minors in Data Science and Mathematics, and graduated in December 2026. I'm originally from Egypt and the first in my family to go to college.

These days I work where data and AI meet. As a Data Engineer / Specialist at Schneider Electric I build Databricks pipelines over 1.3M+ Redshift order records. With Handshake AI I audit LLM output on coding tasks, and before that I spent a summer building solar-performance pipelines at Silicon Ranch.

Here are some technologies I've been working with recently:

  • Python
  • SQL & Spark SQL
  • PySpark · Databricks
  • AWS (S3, Redshift)
  • C / C++
  • C# · .NET
  • TypeScript · Next.js
  • Tableau · Power BI
MTSU logo

Education

Middle Tennessee State University

B.S. Computer Science · Minors in Data Science & Mathematics

  • Honors Program
  • GPA 3.71 / 4.00
  • Graduated Dec 2026
  • Murfreesboro, TN

Coursework: Data Structures & Algorithms, Computer Systems, Software Development, Database Management, Programming Languages, Data Science, Applied Predictive Modeling, Data Analysis, Feature Engineering, Calculus I–III

Steven Gobran smiling, arms crossed, at a restaurant table
Nashville, TN

~/experience

Data Engineer / Specialist @ Schneider Electric

May 2026 – PresentNashville, TN

Joined as a Transformation & Analytics Intern, now a Data Engineer / Specialist.

  • Built scheduled Databricks pipelines in PySpark and Spark SQL that enrich 1.3M+ Redshift order records, validating row counts and order value at every stage so nothing is duplicated downstream in Tableau.
  • Built predictive Python pipelines that forecast headcount demand and utilization for a 100+ person CMO across 16 managers.
  • Traced a 52K-record gap to an upstream status filter by querying information_schema across every warehouse schema, then moved eight fields upstream, which removed two joins and retired a redundant table.
  • Engineering an SFTP-to-MES integration that runs every 30 minutes: Oracle CSV ingestion, SharePoint cross-reference enrichment, deduplication by window ranking, and automated alerts when operator mappings are missing.
  • Databricks
  • PySpark
  • Spark SQL
  • Amazon Redshift
  • Tableau
  • Oracle

~/projects

All repos on GitHub

Full-stack · data pipeline

MTSU GradPath

A degree planner for every MTSU undergraduate major, concentration and minor. Check off what you've taken and it builds a term-by-term plan to graduation that respects prerequisites, while a self-updating pipeline keeps the catalog data current.

  • 223 majors and concentrations, 152 minors and 3,475 courses from the 2026–27 catalog
  • Started as a student team project; I wrote most of the code, including the rewrite from 8 hand-coded majors to all 223 programs
  • Python
  • Flask
  • GitHub Actions
  • Mermaid.js
  • pytest
  • Vercel

Risk analytics · full-stack

Prop Desk

A Flask and PostgreSQL dashboard for managing funded prop-firm futures accounts. Trades from my TradingView bot arrive by webhook, a risk engine works out how much drawdown room each account has left, and Monte Carlo simulations project the odds of passing and getting paid out.

  • 20,000-run bootstrap Monte Carlo that replays every account on one shared trade sequence to capture correlated risk
  • Webhook ingestion routes TradingView/TradersPost events by account and dedupes retries by event ID
  • Python
  • Flask
  • PostgreSQL
  • SQLAlchemy
  • NumPy
  • Chart.js

Backend · distributed services

Open Bank

A multi-service banking platform. An ASP.NET Core API handles auth and PostgreSQL persistence, a separate C++17 microservice validates every transaction and computes its fee over REST before anything commits, and a SignalR hub pushes live balance updates to the dashboard.

  • Money moves inside PostgreSQL transactions with SELECT … FOR UPDATE row locks and an append-only ledger
  • JWT + bcrypt auth; account endpoints are scoped to the caller, admin routes re-check a role claim, and the WebSocket hub re-checks ownership
  • C#
  • ASP.NET Core
  • C++17
  • PostgreSQL
  • SignalR
  • JWT
Lakehouse pipeline diagram Synthetic ETO orders are generated in Python and land in AWS S3, flow through Databricks bronze, silver and gold Delta tables, then load into Amazon Redshift via COPY for Tableau analytics. 01 · generate ETO order data 02 · land AWS S3 databricks · delta lake Bronze raw ingest from S3 Silver refined · PySpark + SQL Gold analytics-ready tables 03 · serve (COPY) Amazon Redshift 04 · analyze Tableau Lakehouse pipeline: ETO order data lands in AWS S3, moves through Databricks bronze, silver and gold Delta tables, then loads into Amazon Redshift via COPY for Tableau. 01 · generate ETO order data · Python 02 · land AWS S3 raw zone databricks · delta lake Bronze Silver · PySpark + SQL Gold · analytics-ready 03 · serve (COPY) Amazon Redshift 04 · analyze Tableau dashboards

Data engineering · cloud

GSC NAM ETO Lakehouse

An end-to-end lakehouse pipeline. Synthetic engineer-to-order (ETO) order data lands in AWS S3, is refined through bronze, silver and gold Delta tables in Databricks with PySpark and SQL, and is then served to Amazon Redshift via COPY for analytics in Tableau.

  • Medallion architecture: raw bronze → refined silver → analytics-ready gold Delta tables
  • Gold tables loaded into Amazon Redshift with COPY and visualized in Tableau
  • Databricks
  • PySpark
  • Spark SQL
  • AWS S3
  • Amazon Redshift
  • Tableau

Data mining · geospatial

Cyclist Behavior from GPS Trajectories

Turns raw smartphone GPS traces from Berlin's SimRa dataset into ride-level behavior features. A Python pipeline cleans 877 ride files, computes speed, stop, turn and braking features with haversine and bearing math, analyzes rider-tagged near misses, and clusters riding styles with K-means. Built with a classmate for a data mining course.

  • 375,741 GPS points across 856 rides (4,936 km), with 408 rider-tagged near-miss incidents
  • Half of all incidents were close passes, and 73% involved a car
  • Python
  • pandas
  • NumPy
  • scikit-learn
  • Matplotlib
  • Jupyter
Sonosus VR Education homepage advertising point-of-care ultrasound and emergency imaging courses

Client project · full-stack

Sonosus VR Education

A course and certification platform I built solo for Dr. Susanna Rudy, who teaches point-of-care ultrasound at Vanderbilt School of Nursing. Learners pay for courses, watch lessons that stay behind a paywall, pass timed exams and earn certificates that anyone can verify, while an admin studio runs courses, students, refunds and emails.

  • Checkout with card, Apple Pay and Google Pay; a purchase unlocks that course's paywalled video lessons
  • Certificates generated when a learner passes the final exam, plus a public page that validates any certificate ID
  • Next.js
  • React
  • TypeScript
  • Tailwind CSS
  • Auth.js
  • Vercel

More things I've built

Concurrent Stock Market Simulator

Systems · C

A multi-threaded C server that simulates a live market for many traders at once: a custom TCP protocol for terminal clients, a hand-written HTTP/1.0 JSON API with a web dashboard, and a writer-priority reader-writer lock built from POSIX semaphores.

  • C
  • pthreads
  • Semaphores
  • TCP
  • HTTP

COOL Compiler Toolchain

Compilers · C++ · coursework

Three CSCI 4160 projects on the course's COOL compiler framework: I wrote the LALR(1) Bison grammar that builds the AST, the type rules for a static checker (subtyping and least-upper-bound typing), and the expression evaluator of a tree-walking interpreter.

  • C++
  • Flex
  • Bison
  • Make

Terminal-Bench Task Repair

AI evaluation

Repaired a broken agent-evaluation task where an AI agent must turn an Apache log into a JSON report. Removed a reference solution that had leaked into the agent's image, pinned the Docker image by digest, cut internet access and replaced an existence-only check with a real pytest verifier.

  • Python
  • pytest
  • Docker
  • TOML

Solar Inverter Fault Monitor

Data engineering · Python

Automated health checks for inverter modules across 21 utility-scale solar sites. A Seeq formula labels each module as faulted, offline, in a comms outage or frozen, only when there's enough sunlight, then groups fault codes into timed periods and exports per-site CSV reports.

  • Python
  • pandas
  • Seeq SPy
  • Jupyter

Boxing Monte Carlo

Simulation · Python

Estimates win, draw and KO odds for any two boxers by simulating 100,000 fights in parallel across CPU cores. Each fight samples record, KO rate, height, reach and weight, with noise scaled to how much data each fighter has. Includes a Flask + Chart.js web UI.

  • Python
  • NumPy
  • multiprocessing
  • Flask
  • Chart.js

JetFriend

AI app · team project

A conversational travel planner built on Flask and GPT-4o. It detects place and itinerary questions, adds curated destination data to the prompt, and has the model reply with styled itinerary cards. Built with two teammates and deployed on Render.

  • Python
  • Flask
  • OpenAI GPT-4o
  • JavaScript

Nashville Crime by ZIP Code

GIS · Python

Assigns 103,218 of 108,287 reported crime locations in Davidson County to its 38 ZIP codes. Dissolved 13,379 Census polygons into ZIP areas in QGIS, counted crimes per area, then rebuilt the workflow in GeoPandas with dissolve and spatial joins.

  • QGIS
  • GeoPandas
  • Python
  • Shapefiles

NYC Subway Homicide Proximity

Spatial SQL

PostGIS analysis that counts homicides within 2,500 ft of each of NYC's 491 subway stations using ST_DWithin on projected geometries, then uses a window function to flag the 195 stations above their borough's average.

  • PostgreSQL
  • PostGIS
  • SQL

Game Glitch Investigator

Applied AI · CodePath

My CodePath AI110 final project: I fixed the bugs in an AI-generated Streamlit guessing game, then added a retrieval tab over a bug-pattern knowledge base, a built-in reliability evaluator and 27 pytest tests.

  • Python
  • Streamlit
  • pytest
  • Matplotlib

Housing Data Dive: Ames, Iowa

Analytics · Python

Exploratory analysis of 1,460 home sales framed as a pricing guide for a real-estate firm. Overall quality (r = 0.79) and living area (r = 0.71) drive price, and neighborhood medians differ by 3.5×. Presented on YouTube as Data Dive 1.

  • Python
  • pandas
  • Matplotlib
  • Jupyter

Concurrent Employee Datastore

Systems · C

A thread-per-client TCP server in C for searching and adding employee records. A writer-priority reader-writer lock lets searches by name, ZIP or salary run in parallel while writes get exclusive access.

  • C
  • pthreads
  • Semaphores
  • TCP

BugHound Guardrails

AI agents · CodePath

Added reliability guardrails to a code-fixing agent that proposes patches with Gemini: an AST syntax check that blocks broken fixes from auto-applying, a penalty for bloated patches, and a fallback when the model doesn't return clean JSON.

  • Python
  • Gemini API
  • Streamlit
  • pytest

I trade Micro Nasdaq futures on funded prop-firm accounts with a bot I wrote, and I built the software that runs the desk around it. Prop Desk tracks every account's risk, and a trading agent I built recalculates position sizes each day, updates my TradingView alerts and sends the results to my phone.

The agent's report on my phone: the day's result, payout odds from 3,000 simulated paths, and every alert it resized and verified for Monday.

Pine Script · MNQ 15-minute · trades my funded accounts

Sniper Range Bot V2

An automated range-breakout strategy for Micro Nasdaq-100 futures. Position size comes from a fixed risk per trade, every trade uses a 1:1 reward-to-risk, and a weekday filter trades only the direction that has held up historically on that day. It's built to stay inside prop-firm drawdown and consistency rules.

TradingView backtest · Mar 31 – Sep 25, 2026

Win rate75.86%66 of 87 trades
Profit factor3.16gross profit ÷ gross loss
Net P&L+$92.5K+185% on a 50K backtest account
Max drawdown7.11%peak to trough, per TradingView
Backtested cumulative P&L Backtest · closed trades, Mar 31 – Sep 25, 2026 · 50K account

Line chart of cumulative profit across 87 backtested trades, rising from $0 on March 31 to $92,538 on September 25, 2026. Use the left and right arrow keys to step through trades. A full table follows the chart.

  • MonLong only
  • TueBoth
  • WedLong only
  • ThuShort only
  • FriBoth

TradingView backtest, MNQ 15m, Mar 31 – Sep 25, 2026. Past performance does not guarantee future results. Not financial advice.

View all 87 backtest trades as a table
#Exit dateSideExitP&LCumulative
1Apr 1, 2026LongLong TP+$2,070+$2,070
2Apr 2, 2026ShortShort SL−$2,070$0
3Apr 3, 2026ShortShort TP+$2,088+$2,088
4Apr 7, 2026ShortShort SL−$2,040+$48
5Apr 8, 2026LongLong TP+$2,062+$2,110
6Apr 10, 2026ShortShort TP+$2,079+$4,190
7Apr 13, 2026LongLong TP+$2,092+$6,282
8Apr 17, 2026LongLong TP+$2,068+$8,350
9Apr 21, 2026LongLong SL−$2,046+$6,304
10Apr 22, 2026LongLong TP+$1,998+$8,302
11Apr 23, 2026ShortShort SL−$2,100+$6,202
12Apr 24, 2026LongLong TP+$2,072+$8,274
13Apr 28, 2026ShortShort TP+$2,090+$10,364
14Apr 30, 2026ShortShort SL−$1,992+$8,372
15May 1, 2026ShortShort TP+$2,090+$10,462
16May 5, 2026LongLong TP+$2,100+$12,562
17May 6, 2026LongLong SL−$2,067+$10,494
18May 8, 2026LongLong TP+$2,070+$12,564
19May 12, 2026ShortShort TP+$2,088+$14,652
20May 13, 2026LongLong TP+$2,016+$16,668
21May 15, 2026ShortShort TP+$1,926+$18,594
22May 19, 2026LongLong TP+$2,090+$20,684
23May 20, 2026LongLong TP+$2,016+$22,700
24May 21, 2026ShortShort TP+$2,076+$24,776
25May 22, 2026ShortShort TP+$2,080+$26,856
26May 25, 2026LongLong SL−$2,094+$24,762
27May 26, 2026LongLong SL−$2,033+$22,729
28May 29, 2026LongLong TP+$2,086+$24,816
29Jun 2, 2026LongLong TP+$2,062+$26,878
30Jun 3, 2026LongLong TP+$2,060+$28,938
31Jun 5, 2026ShortShort TP+$2,024+$30,962
32Jun 9, 2026LongLong TP+$2,030+$32,992
33Jun 10, 2026LongLong SL−$2,018+$30,974
34Jun 12, 2026ShortShort TP+$1,989+$32,962
35Jun 15, 2026LongLong SL−$2,057+$30,906
36Jun 16, 2026ShortShort SL−$2,030+$28,876
37Jun 17, 2026LongLong TP+$2,046+$30,922
38Jun 19, 2026ShortShort SL−$2,082+$28,840
39Jun 22, 2026LongLong TP+$1,994+$30,834
40Jun 23, 2026ShortShort TP+$1,962+$32,796
41Jun 24, 2026LongLong SL−$1,998+$30,798
42Jun 26, 2026LongLong TP+$2,043+$32,840
43Jun 29, 2026LongLong TP+$2,076+$34,916
44Jun 30, 2026ShortShort TP+$2,016+$36,932
45Jul 2, 2026ShortShort TP+$2,010+$38,942
46Jul 3, 2026LongLong TP+$1,962+$40,904
47Jul 7, 2026LongLong TP+$2,100+$43,004
48Jul 10, 2026ShortShort TP+$2,025+$45,030
49Jul 14, 2026LongLong TP+$2,061+$47,090
50Jul 17, 2026ShortShort SL−$2,082+$45,008
51Jul 21, 2026LongLong TP+$2,062+$47,070
52Jul 24, 2026ShortShort TP+$2,090+$49,160
53Jul 27, 2026LongLong TP+$2,040+$51,200
54Jul 28, 2026ShortShort TP+$2,040+$53,240
55Jul 29, 2026LongLong SL−$2,082+$51,158
56Jul 31, 2026LongLong TP+$2,085+$53,242
57Aug 4, 2026LongLong TP+$2,079+$55,322
58Aug 5, 2026LongLong TP+$2,079+$57,400
59Aug 6, 2026ShortShort TP+$1,980+$59,380
60Aug 7, 2026LongLong TP+$2,062+$61,442
61Aug 10, 2026LongLong TP+$2,048+$63,490
62Aug 11, 2026ShortShort TP+$2,025+$65,514
63Aug 12, 2026LongLong TP+$2,091+$67,606
64Aug 14, 2026LongLong TP+$2,092+$69,698
65Aug 17, 2026LongLong TP+$2,052+$71,750
66Aug 18, 2026ShortShort TP+$2,054+$73,804
67Aug 21, 2026LongLong TP+$2,082+$75,886
68Aug 25, 2026LongLong TP+$2,040+$77,926
69Aug 27, 2026ShortShort TP+$2,058+$79,984
70Aug 28, 2026ShortShort SL−$2,079+$77,906
71Aug 31, 2026LongLong TP+$2,032+$79,938
72Sep 1, 2026LongLong TP+$2,091+$82,028
73Sep 2, 2026LongLong SL−$2,042+$79,986
74Sep 3, 2026ShortShort SL−$1,904+$78,082
75Sep 4, 2026LongLong SL−$2,026+$76,056
76Sep 7, 2026LongLong TP+$2,064+$78,120
77Sep 8, 2026ShortShort TP+$2,074+$80,194
78Sep 11, 2026LongLong TP+$1,981+$82,175
79Sep 15, 2026LongLong SL−$2,032+$80,144
80Sep 16, 2026LongLong TP+$2,058+$82,202
81Sep 17, 2026ShortShort TP+$2,100+$84,302
82Sep 18, 2026LongLong TP+$2,060+$86,362
83Sep 21, 2026LongLong TP+$2,079+$88,440
84Sep 22, 2026ShortShort SL−$2,052+$86,388
85Sep 23, 2026LongLong TP+$2,088+$88,476
86Sep 24, 2026ShortShort TP+$2,082+$90,558
87Sep 25, 2026LongLong TP+$1,980+$92,538

How the desk runs itself

Prop Desk code
  1. 01

    Signal

    Sniper Range Bot V2 spots a range breakout on the 15-minute MNQ chart and fires a TradingView alert with a 1:1 take profit and stop.

  2. 02

    Execute

    TradersPost turns that alert into an order on every connected prop-firm and personal account, so no one clicks anything.

  3. 03

    Track

    Each closed trade posts its P&L to Prop Desk, which recalculates every account's room, survivable losses and trailing floor.

  4. 04

    Recalculate

    My agent sets the next session's size for each account, updates and verifies the TradingView alerts, reruns the Monte Carlo odds and emails me the report.

More from the trading desk

~/research

Title page of the paper Security Audit of WordPress 4.7.0 Open PDF
Published April 2026 · Zenodo With Mena Ramzy · CSCI 4710

Security Audit of WordPress 4.7.0: A Comparative Evaluation of SAST and DAST Tools Against a Known CVE Record

We audited WordPress 4.7.0, a release with well-documented critical CVEs, using one static tool (Semgrep) and one dynamic tool (WPScan) in an isolated Docker Compose lab, then mapped every finding to CWE and the OWASP Top 10. WPScan confirmed 114 known CVEs with zero false positives, while Semgrep raised 232 code-level flags at about a 2.5% true-positive rate. The two tools cover different ground, and neither is enough alone.

  • Python
  • Docker
  • Semgrep
  • WPScan
  • OWASP Top 10
  • CWE

Certifications 20 credentials

Also completed

  • Cloud Technical EssentialsAWS
  • Intro to SQLKaggle
  • PythonKaggle
  • Advanced SQLKaggle
  • Introduction to Generative AIGoogle
  • Claude 101Anthropic

~/videos

YouTube channel
Steven holding a Kodak camera up to his eye

@StevenGG488

Walkthroughs of my projects and data dives, from planning to code.

Subscribe