Trevor Santiago

ABOUT

Applied AI engineer

I'm Trevor Santiago, a data scientist at ALLDATA building AI applications and data workflows. My recent work includes Gemini-powered services, semantic retrieval, and tools that help subject-matter experts reuse prior edits and review AI-generated drafts.

I like taking an unclear problem through research, prototyping, and implementation. At ALLDATA, that has meant working directly with subject-matter experts, building FastAPI services on Google Cloud, and checking model output against the cases those experts care about.

Outside work, I built NormFlow, a text-normalization workbench that combines exact matching, semantic retrieval, and optional LLM suggestions with batch processing and human review.

Sacramento, California · English, fluent/native

US citizen, no sponsorship required.

US remote work preferred. Open to relocation for hybrid work in the San Francisco Bay Area or San Diego.

Read the complete Profile

EXPERIENCE

Experience

A record of roles, responsibilities, and the work behind them.

November 2021 to present · ALLDATA · Elk Grove, California

Data Scientist

Build AI services, human review workflows, and cloud data pipelines.

Built the Repair Trends AI service with FastAPI, Cloud Run, Gemini through Vertex AI, and an AlloyDB cache that regenerates results when upstream component inputs change. The beta/staging rollout is complete.
Independently built a Probable Cause Data editing system that fills text edits from a library of previously applied edits using progressive matching. When no match is found, it uses a separately fine-tuned Flan-T5 Small model to generate a new draft edit.
Fine-tuned the generation model with QLoRA and evaluated it with BLEU separately from the editing workflow. The primary SME estimated that the overall system reduced total editing effort by at least 50%.
Owned the Known Fixes pipeline for billions of dealership repair-order records, using Gemini with web-search grounding to standardize free-text part descriptions before ACES mapping.
Combined repair-order data with SME-curated Probable Causes, applied related-vehicle rules, and automated monthly Known Fixes publishing with BigQuery, Cloud Run Jobs, Airflow, and AlloyDB. Known Fixes is a core feature of Diagnostic Intelligence, which reached 1,700+ subscriptions in its first three months.

Read this experience in the Profile

January 2021 to August 2021 · New York Mets

Data Science Intern

Built predictive-model components and venue features for a larger outfield-alignment system.

Owned an XGBoost hit-outcome classifier within a multi-model system designed to maximize expected outs for batter and pitcher matchups.
Trained and evaluated the classifier using separate historical training, evaluation, and held-out test sets.
Standardized wall geometry for every MLB venue with interpolation and spline fitting, estimating distance from home plate at one-degree increments.

Read this experience in the Profile

PROJECT EVIDENCE

NormFlow

Independent project

A text-normalization workbench with exact matching, semantic retrieval, optional LLM fallback, and human review.

Python
FastAPI
TypeScript
SQLite
FAISS
Semantic retrieval
Human review

I built and shipped NormFlow using Python, FastAPI, TypeScript, SQLite, and FAISS. It combines exact matching and semantic retrieval with optional LLM fallback.

Durable batch, review, and export workflows keep human review and correction within the process of turning raw text into reusable normalized data.

View on GitHub Open the project page Find it in the Profile

PROJECT EVIDENCE

AI-assisted PCD editing

ALLDATA

An editing workflow that fills from previously applied edits, then uses a separately fine-tuned model to generate new draft edits when matching finds no reusable edit.

Python
Semantic retrieval
Flan-T5
QLoRA
BLEU
Cloud Run Jobs
Cloud Storage

Probable Cause Data editing turns parsed OEM repair text into curated probable causes. Subject-matter experts used an Excel macro and a library mapping source text to previously applied edits. Its limited matching left unmatched text for manual drafting.

I built a workflow that looks for each input in that library, progressing through exact, case-insensitive, whitespace-normalized, formatting-normalized, and semantic matching. A match fills the corresponding prior edit into the output. Normalization preserves symbols with automotive meaning and exposes conflicting library mappings for expert review.

Inputs without a reusable match go to a separate generation step. This step uses a fine-tuned Flan-T5 Small model to generate new draft edits.

Matching and generation run as separate Cloud Run Jobs connected by Cloud Storage upload events. I also built a simple upload/download UI for non-technical experts. I designed and implemented the system independently apart from the UI's Kubernetes deployment, which a DevOps engineer owned.

The primary subject-matter expert estimated at least a 50% reduction in total editing effort.

Open the project page Find it in the Profile

PROJECT EVIDENCE

Repair Trends AI service

ALLDATA

A Gemini-powered service that supplements repair-order observations with component-failure suggestions for a vehicle and mileage range.

Python
FastAPI
Gemini
Vertex AI
Cloud Run
AlloyDB
Caching

Repair Trends helps technicians identify components likely to fail and cause an unscheduled repair within a requested mileage range. The upstream repair-order data had coverage gaps and could include scheduled-wear or collision-related replacements.

I independently built and deployed the AI service using FastAPI on Cloud Run and Gemini through Vertex AI. Teammates owned the upstream cleaning pipeline and querying API.

The service uses aggregated repair-order observations as grounding to rerank components, revise probabilities, and suggest missing components. It accounts for scheduled-wear and collision-related biases and generates suggestions when the upstream component list is empty.

An AlloyDB cache reuses responses for repeated vehicle, mileage, and component inputs. When the upstream component list changes, the service regenerates and replaces the cached response.

Direct data covered 80 of 100 expert-selected vehicle and mileage test cases. The service produced AI-assisted output for all 100 when available, including the 20 without underlying observations. Subject-matter experts reviewed every case. This is test-set coverage, not an accuracy measurement.

The beta/staging rollout is complete as of October 2026.

Open the project page Find it in the Profile

PROJECT ARCHIVE

Project archive

More projects from my work and independent builds.

Known Fixes

ALLDATA

A repair-intelligence pipeline combining dealership repair orders and expert-curated probable causes to rank parts associated with diagnostic trouble codes.

Python
SQL
Gemini
BigQuery
Cloud Run Jobs
Airflow
AlloyDB

Read project details

Mets outfield-alignment modeling

New York Mets

An XGBoost hit-outcome classifier and standardized ballpark geometry for matchup-specific outfield alignment.

Python
XGBoost
Classification
Feature engineering
Interpolation

Read project details

Find a Fix data pipeline

ALLDATA · Retired project

BigQuery and Dataform pipelines over millions of retail-sales, lookup, and scan records for a legacy repair-ranking product.

SQL
BigQuery
Dataform
Cloud Run
Cloud Storage
Solr

Read project details

View the project archive

SKILLS & EDUCATION

Skills & education

Tools, techniques, and a few foundations.

Skills

Python
SQL
LLM applications
Semantic retrieval
Prompt engineering
Data pipelines
FastAPI
Flask
Docker
TypeScript
SQLite
scikit-learn
XGBoost
Pandas
NumPy
FAISS
AlloyDB / PostgreSQL
BigQuery
Dataform
Airflow
Vertex AI
Cloud Run
Cloud Functions
Cloud Storage
Solr
Claude Code
Codex
Vercel

Education

August 2020 to August 2021

M.S. in Data Science

University of San Francisco

Selected coursework: machine learning, deep learning, databases, distributed computing, and design of experiments.

July 2016 to June 2020

B.S. in Mathematical Sciences

University of California, Santa Barbara

Minor in Statistical Science. Selected coursework: linear algebra, probability, stochastic processes, operations research, and linear regression.

Certifications

Fundamentals of LLMsHugging Face · June 2025
LangChain Chat with Your DataDeepLearning.AI · May 2025
LangChain for LLM Application DevelopmentDeepLearning.AI · May 2025
Free Data Engineering Bootcamp CertificateDataExpert.io · January 2025
Google Cloud Fundamentals: Core InfrastructureGoogle · November 2023

Recognition

Extra Miler of the YearALLDATA · 2023

Read skills and education in the Profile

AI Q&A

Ask about Trevor

UNDER CONSTRUCTION

AI Q&A is on the way.

I'm building and testing a Q&A that answers questions about my work and links to the supporting evidence.

For now, you can read the projects and experience directly or get in touch.

Browse projects Get in touch

THE GARDEN HALL

The hall couldn't load.

Trevor Santiago

Experience

Data Scientist

Data Science Intern

NormFlow

AI-assisted PCD editing

Repair Trends AI service

Project archive

Known Fixes

Mets outfield-alignment modeling

Find a Fix data pipeline

Skills & education

Skills

Education

M.S. in Data Science

B.S. in Mathematical Sciences

Certifications

Recognition

Ask about Trevor

AI Q&A is on the way.