Cole Hoffman

Forward Deployed Applied AI Engineer

High ownership execution across ambiguous, high stakes problems spanning applied AI, product delivery, stakeholder management, and adoption.

Best fit for teams that need a builder who can find the wedge, work directly with users, ship quickly, and turn early momentum into durable systems.

Best fit

Founding engineer, forward deployed engineer, applied AI lead, and quant ML systems roles.

Forward Deployed / Founding Engineer

For teams that need founder-style execution, stakeholder ownership, product judgment, and technical range.

Open PDF →

Quant / ML Systems

For roles centered on research systems, quantitative modeling, MLOps, and investment infrastructure.

Open PDF →

Ad Hoc Consulting

Engagement pricing, scope tiers, and terms for consulting and contract work.

Open PDF →

Selected Impact

$50M+

Annual alpha opportunity from MBS prepayment modeling

28K

Products forecasted across 36 brands

<14 days

Cloneable systematic investment pipeline delivered

300x

Backtesting workflow acceleration

5

Concurrent internal client engagements

3

Business units supported as an SME

Signature Systems

MBS Prepayment Modeling

Problem
The firm was paying for YieldBook licenses to model conditional prepayment rates.

Solution
Built TB scale data pipelines and ran experiments across CatBoost, XGBoost, and TensorFlow.

Outcome
Delivered a 20 model ensemble that beat the incumbent vendor in 18 months, tied to a $50M+ annual alpha opportunity.

Why it matters
Replaced a paid external dependency with a stronger internal modeling capability tied directly to trading decisions and alpha generation.

CatBoostXGBoostTensorFlowMLflowSHAP

Fund Cash Flow Forecasting

Problem
A custom TensorFlow and Keras model lived across five notebooks with a five day training cycle.

Solution
Re architected the workflow into two SageMaker pipelines for production inference and distributed backtesting, with processing rewritten in Polars.

Outcome
Ran 360 models in parallel, completed a full 1996 to 2024 backtest in four days, and supported forecasts across 28K products and 36 brands.

Why it matters
Converted a slow research workflow into production grade infrastructure that improved speed, scale, and decision support.

TensorFlowKerasSageMakerPolars

SEC 10-K RAG Platform

Problem
Portfolio managers needed a faster way to work across SEC filings by company, year, quarter, and section.

Solution
Built end to end EDGAR ingestion, SEC XML parsing, vector storage in Pinecone, and LLM powered retrieval with intelligent query decomposition.

Outcome
Handed the platform off to the ML engineering team for production scaling and drove adoption across 5+ teams.

Why it matters
Turned unstructured filings into a usable research workflow and created reusable infrastructure for broader internal adoption.

RAGPineconeEDGARInternal GTM

GenAI Interpretability Layer

Problem
Researchers were losing days to manual experiment interpretation and hypothesis review.

Solution
Built an agentic system that ingests evaluation metrics, SHAP outputs, pipeline code, and researcher axioms, then produces executive ready reports and structured review workflows.

Outcome
Improved research velocity and was adopted across a 14 person research team.

Why it matters
Compressed manual interpretation work into a repeatable system that improved research speed, clarity, and team-wide consistency.

AWS BedrockAgentic workflowsInterpretabilityResearch enablement

Career Arc

From data foundations to forward deployed applied AI leadership.

09/2024 to Present

Senior Applied AI Engineer / Tech Lead

Global Investment Systems, Vanguard

Now leading high ambiguity applied AI work across research, product delivery, stakeholder management, and team execution.

08/2023 to 09/2024

Applied AI Engineer

Global Investment Systems, Vanguard

Shifted into internal consulting, turning research prototypes into production AI systems and stakeholder facing products.

12/2022 to 08/2023

Full-Stack Engineer

Equity Investment Systems, Vanguard

Expanded into product delivery with analytics platforms and benchmark tooling for portfolio managers.

04/2022 to 11/2022

Machine Learning Engineer

Enterprise Advice, Vanguard

Moved into applied ML with personalization, tax optimization, and early MLOps foundations.

08/2021 to 04/2022

Data Engineer

Chief Data Analytics Office, Vanguard

Built the data foundations: enterprise ETL, metadata systems, and compliance workflows at scale.

What I'm Building

Machiavelliam

AI poker coaching app (Flutter + GenAI). Trains decision-making under uncertainty, not memorization.

GitHub →

Elev8 Prototype

Polymarket + Reddit social platform. Building with a cofounder. Link to the barebones MVP.

GitHub →

gis-phl

A repository for exploring geospatial applications to help the city of Philadelphia.

GitHub →

Core Toolbox

Languages & Frameworks

PythonSQLRTypeScriptNext.js

ML/AI

TensorFlowCatBoostXGBoostscikit-learnSHAPTime series forecastingSurvival modeling

Domain

Factor modelingPortfolio optimizationRisk modeling

Cloud & Infrastructure

SageMakerAWS BedrockMLflowAzure OpenAIRAGAgentic workflowsPineconeQdrant

Data

PolarsSparkEMRGlueAthenaS3PostgresGeoPandas

DevOps

DockerTerraformCI/CD

Creative

Unreal EngineBlender