Devanssh P
Dallal

I put AI to work on the factory floor where Airbus builds the A320.

Forward Deployed AI & Analytics Engineer at Airbus in Mobile, Alabama. Mathematics graduate of UT Austin.

The autopilot learns by the cross-entropy method: 40 candidates a generation, each flown through the same gusts, the best 8 kept. The gusts are an Ornstein-Uhlenbeck process. The four lights by the runway are a PAPI: two red, two white means on the glide path.

Howdy, I’m Devanssh.

I studied mathematics, then went looking for the place where it gets its hands dirty. I found it on an aircraft assembly line.

At Airbus I sit between the people who build A320s and the data that describes how they do it. I deploy chatbots, monitoring tools and incident-routing agents, rebuild pipelines in Palantir Foundry, and teach colleagues to use generative AI in their own work.

Before that I moved petabytes to Azure, tuned Snowflake, wrote test tooling for medical devices and analysed gene expression at Harvard Medical School. On my own time I ship apps and websites for people who need them.

Now
Airbus, Digital for Operations
Based in
Mobile, Alabama
Studied
B.S. Mathematics, The University of Texas at Austin
Works in
Python, PySpark, SQL, TypeScript
Speaks
English, Hindi, Haryanvi, Punjabi, Latin, some Spanish and French

The route so far

Four cities on the map, then every stop in order, most recent first.

Map of the United States marking the four cities I have worked in: Boston, Austin, Houston and Mobile, joined in that order. Boston Austin Houston Mobile
  1. 2026Mobile, AL

    Airbus, Forward Deployed AI & Analytics Engineer

    January 2026 to now

    AI agents, pipelines and data products for the A320 final assembly line. See what I did

  2. 2024Mobile, AL

    Airbus Global Graduate Program

    Straight from graduation at UT Austin

    Three rotations through the digital side of an aircraft factory. See what I did

  3. 2023Houston, TX

    Daikin

    Data Engineering Intern

    Two Snowflake instances folded into one. See what I did

  4. 2022Austin, TX

    International Biomedical

    Software Engineering Intern

    Test tooling for incubators that carry newborns between hospitals. See what I did

    The same spring I was Recruitment Chair for Tau Kappa Epsilon: a five-week rush, 14 new members, 17 events with 2,000 people through the door, and recruitment costs down 10%.

  5. 2021Austin, TX

    Dover Fueling Solutions

    Cloud Engineering Intern

    Two petabytes of archives, moved to Azure. See what I did

  6. 2020Austin, TX

    The University of Texas at Austin

    B.S. Mathematics, with a certificate in Scientific Computation and Data Sciences

    Arrived mid-pandemic and joined Tau Kappa Epsilon. Spent freshman year helping two student startups grow: MatchUT and Gumdrop.

  7. 2019Boston, MA

    Harvard Medical School

    Bioinformatics Analyst

    A summer of cancer and microbiome research while still in high school. See what I did

  8. 2017Flower Mound, TX

    Marcus High School

    2017 to 2020, high school diploma

    Sophomore year through graduation. The summer before senior year went to a research lab in Boston instead of a beach.

  9. 2016Marietta, GA

    Walton High School

    2016 to 2017

    Freshman year, before the move to Texas.

Where I’ve worked

Most recent first. Every label under a job is color-coded by what kind of thing it is, and each internship opens into a longer write-up.

  • Languages
  • Platforms and cloud
  • Software and tools
  • Skills

January 2026 to now

Forward Deployed AI & Analytics Engineer

Airbus, Mobile, Alabama

What I did

  • Deploy AI chatbots, performance-monitoring tools and machine-learning agents that route production incidents on the Mobile final assembly line.
  • Rebuilt the A320 capacity reporting pipeline as PySpark transforms in Palantir Foundry, making it faster and more accurate.
  • Fixed 55 bugs across manufacturing data products, hardening the pipelines other teams depend on.
  • Product Owner for moving an A350 parts plant’s data products onto standard Airbus systems, and for an asset-tracking product.
  • Help run the AI Innovation Guild, serve as a Gemini Ambassador, and lead the STEAM Squad that takes hands-on science into local schools. Top-three project finalist at the Airbus Leadership Academy in 2025 and 2026.

What I used

  • Language: Python
  • Language: PySpark
  • Language: SQL
  • Platform: Palantir Foundry
  • Platform: Airbus Skywise
  • Platform: Google Cloud
  • Platform: SAP
  • Tool: Gemini
  • Tool: Polars
  • Skill: AI agents
  • Skill: Machine learning
  • Skill: Data pipelines
  • Skill: Incident routing
  • Skill: Product ownership
  • Skill: Teaching AI

June 2024 to December 2025

Airbus Global Graduate Program

Airbus, Mobile, Alabama

What I did

Three rotations through the digital side of an aircraft factory: Industrial Digital Transformation, MES Transformation, then Performance and Competitiveness Management.

  • Built machine-learning incident-management tools on Palantir Foundry that cut the time to resolve a production incident by 28%.
  • Diagnosed root causes in the SAP ME manufacturing execution system rollout and presented the findings to its steering committee.
  • Measured deployment performance for the Columbus final assembly line rollouts, so progress could be benchmarked.
Mean time to resolve a production incident
Before6 days
After4.3 days

What I used

  • Language: Python
  • Language: SQL
  • Platform: Palantir Foundry
  • Platform: Airbus Skywise
  • Tool: SAP ME
  • Skill: Machine learning
  • Skill: Root-cause analysis
  • Skill: Performance benchmarking
  • Skill: Manufacturing systems
  • Skill: Steering-committee reporting
  • Skill: SAFe

June to August 2023

Data Engineering Intern

Daikin, Houston, Texas

What I did

  • Migrated the marketing Snowflake instance into the main one with better ETL, cutting query time by 5%.
  • Remapped data in the main Snowflake instance, saving about $15,000 a year and making storage more accessible and secure.
  • Built Tableau views of commercial and residential sales so dealers could see their market.

What I used

  • Language: SQL
  • Platform: Snowflake
  • Tool: Tableau
  • Skill: ETL
  • Skill: Data migration
  • Skill: Query optimization
  • Skill: Data visualization

Read the full Daikin write-up

May to July 2022

Software Engineering Intern

International Biomedical, Austin, Texas

What I did

  • Wrote a data-parsing application that converts machine-generated CSV files into structured JSON for testing neonatal incubators.
  • Designed an automated testing framework that sped up manufacturing analysis.
  • Wrote the scripts in Python and integrated them with existing systems in C++, inside Docker so every environment matched.
  • Worked in GitFlow on GitHub, with Jira for tracking across teams.

What I used

  • Language: Python
  • Language: C++
  • Tool: Docker
  • Tool: GitHub
  • Tool: Jira
  • Skill: Test automation
  • Skill: Data parsing
  • Skill: GitFlow

Read the full International Biomedical write-up

June to August 2021

Cloud Engineering Intern

Dover Fueling Solutions, Austin, Texas

What I did

  • Carried out the transfer of two petabytes of company archives from a legacy SAP server to Microsoft Azure.
  • Devised an algorithm that cleans and wrangles gas-station transaction data into business intelligence reports.

What I used

  • Language: C#
  • Platform: Microsoft Azure
  • Platform: SAP
  • Skill: Cloud migration
  • Skill: Data wrangling
  • Skill: Business intelligence

Read the full Dover write-up

January to October 2021

Strategy Specialist at UT Austin

Gumdrop, a campus group-chat startup

What I did

  • Led marketing at UT Austin for a Los Angeles startup that connects incoming freshmen before they arrive.
  • Developed the diagnostic test that sorts new users into the right groups.
  • Beta-tested the app and fed feature recommendations back to the team.
  • Organized and hosted events to keep people active on the platform.

What I used

  • Skill: Campus marketing
  • Skill: Quiz design
  • Skill: Beta testing
  • Skill: Product feedback
  • Skill: Event hosting

December 2020

Marketing Ambassador

MatchUT, a matchmaking site for UT Austin students

What I did

  • Ran word-of-mouth and one-on-one marketing that brought 7,700 students to sign up in seven days, about 20% of the student body.
  • Helped drive more than 20,000 website views, reaching roughly half of all students.
  • Helped make MatchUT trend on Twitter during a semester when campus social life had moved online.

What I used

  • Skill: Word-of-mouth marketing
  • Skill: Campus outreach
  • Skill: Viral growth

May to August 2019

Bioinformatics Analyst

Harvard Medical School, Boston, Massachusetts

What I did

Worked with Bhasin Labs and Beth Israel Deaconess Medical Center on cancer and microbiome research.

  • Analysed RNA sequencing data from rat colon cells, 36,000 genes per cell, for a study on treating colon cancer with Withania somnifera, using heat maps, box plots and pathway analysis.
  • Investigated the human gut microbiome to support antibiotic development.
  • Built the figures for a paper on applying a rheumatoid arthritis medication to systemic lupus.

What I used

  • Language: R
  • Language: MATLAB
  • Language: Perl
  • Tool: Ingenuity Pathway Analysis
  • Skill: RNA sequencing analysis
  • Skill: Bioinformatics
  • Skill: Statistics
  • Skill: Scientific figures

Read the full Harvard write-up

Things I’ve built

Some at work, most on evenings and weekends.

Opening Pandora’s Box

A generative AI article series I write for Airbus colleagues, read by more than 28,000 people across the company. Every article ships with a custom Gemini Gem or tool, so readers can try the idea in their own work that afternoon.

  • Platform: Google Cloud
  • Tool: Gemini
  • Skill: Technical writing
  • Skill: Teaching AI

Post-Pigeon mail drone

A fixed-wing cargo drone for the stretch of the mail’s journey between a regional sorting hub and a small-town post office, which is done by truck today. It flies low over the post office, releases a pod of parcels from its belly without landing, and heads home.

The airframe is written as Python CAD code in five stages: outer shape, mail pod, latch, internal structure, then the duct and tailhook. A vortex-lattice study checks stability and a mass table tracks every part. An investor pitch covers the business case.

  • Language: Python
  • Tool: build123d
  • Skill: Parametric CAD
  • Skill: Aerodynamics
  • Skill: Mass and balance
  • Skill: Business case

Small language models

Language models small enough to run on my own machine, built end to end instead of called through an API: preparing the text, training the model, then testing what it writes.

Building one shows what a hosted model hides. I can explain how a model picks its next word, judge when a small model is enough for a job, and tell when fine-tuning is worth the effort over better prompts or retrieval. It is the hands-on side of what I teach at Airbus: part 4 of Opening Pandora’s Box covers fine-tuning and LoRA adapters.

  • Language: Python
  • Tool: PyTorch
  • Skill: Small language models
  • Skill: Model training
  • Skill: Fine-tuning
  • Skill: Model evaluation

My own RAG pipeline

A retrieval-augmented generation system I built from its parts instead of using an off-the-shelf product. Documents are split into passages and turned into embeddings. A question pulls back the closest passages by cosine similarity, and the model answers from those instead of from memory.

This is the pattern behind most assistants that answer from a company’s own documents, and the main defence against a model making things up. Having built each stage, I know where these systems go wrong: bad chunking, weak retrieval, or a prompt that lets the model ignore its sources. Part 2 of Opening Pandora’s Box teaches the same idea to Airbus colleagues.

  • Language: Python
  • Skill: Retrieval-augmented generation
  • Skill: Embeddings
  • Skill: Vector search
  • Skill: Grounded answers

Healing Path

A numerology report generator for a professional numerologist. I turned her course material and report format into a tested calculation engine, so every report follows her method exactly. It runs entirely in the visitor’s browser and stores nothing.

  • Language: TypeScript
  • Platform: Vercel
  • Tool: Next.js
  • Tool: Zod
  • Tool: Vitest
  • Tool: Claude Code
  • Skill: Spec-first development

ShiftBank

A phone app for bartenders, servers and gig workers whose income changes every week. Log a shift in one tap, see what you really earn per hour, set aside the right amount for tax, and compare tips with anonymous averages from the same venue.

  • Language: TypeScript
  • Platform: Supabase
  • Tool: Expo
  • Tool: React Native
  • Tool: RevenueCat
  • Skill: Offline-first sync

Dalal Digital Studios

My studio for small businesses around Mobile that want less busywork: websites, automation and AI set up for people who do not have an IT department. The site includes a free health check that scores how digital a business already is.

  • Language: TypeScript
  • Platform: Supabase
  • Platform: Vercel
  • Tool: Next.js
  • Tool: Tailwind
  • Tool: Resend

Gulf Coast Hub

A local directory and events calendar for the Alabama Gulf Coast, starting in Mobile. Businesses and events are submitted by locals and reviewed before they appear.

  • Language: TypeScript
  • Platform: Supabase
  • Tool: Next.js
  • Tool: Prisma
  • Tool: Stripe

Downtown Car Care

A staff dashboard for a local auto repair shop: which car is in which bay, what each job needs, and which parts are running low. A customer side for booking and repair status comes next.

  • Language: Dart
  • Platform: Supabase
  • Tool: Flutter
  • Tool: Riverpod

Education and toolkit

2020 to 2024

The University of Texas at Austin

Bachelor of Science in Mathematics. Certificate in Scientific Computation and Data Sciences. GPA 3.63.

Languages
  • Python
  • PySpark
  • SQL
  • TypeScript
  • JavaScript
  • Java
  • C++
  • R
  • MATLAB
Platforms and cloud
  • Palantir Foundry
  • Airbus Skywise
  • Google Cloud
  • Snowflake
  • Databricks
  • AWS
  • Azure
  • SAP
Software and tools
  • Gemini
  • Claude Code
  • Model Context Protocol
  • Pandas
  • Polars
  • PyTorch
  • scikit-learn
  • Tableau
  • Docker
Skills
  • Machine learning
  • AI agents
  • Data pipelines
  • Data science
  • Product ownership
  • Teaching AI
Certifications
Airbus Skywise Technical Leader, Certified SAFe 6 Practitioner, Machine Learning with Python, Materials Data Sciences and Informatics, Intelligent Machining, Introduction to User Experience Design
Off the clock
Formula 1 (forza Ferrari), martial arts, ukulele, beatboxing, Led Zeppelin, Longhorn sports and the Premier League

Into AI, aircraft, F1 or good music? Say howdy.