Forward Deployed Engineer
I embed with customers, find the real workflow problem, and ship working software against it. 8+ years of enterprise technical sales and solution engineering across storage, cloud, and SaaS, now paired with hands-on Python, FastAPI, and applied-AI engineering.
Forward deployed roles need someone who can walk into a customer's environment, understand their actual problem, and build software against it on a short feedback loop. That's the combination I bring — a decade of enterprise technology sales and solution engineering, now backed by hands-on software and AI/ML delivery.
8+ years running discovery, demos, and competitive positioning inside enterprise accounts at HPE, Sirius Computer Solutions, and BILL.com. I know how to sit in a room with a skeptical technical buyer, ask the right questions, and figure out what actually has to be true for a deal — or a deployment — to work.
Over the past year I taught myself to build, not just sell: Python, FastAPI, REST APIs, multimodal LLM integration, and applied ML — shipped as working prototypes, not slideware. Every project below is code I wrote and pushed myself.
From a wearable AI assistant to MLOps reproducibility pipelines to telecom churn models, I go from zero context to a working demo fast. That is the muscle forward deployed work actually requires: show up, learn the domain, build the thing.
A blend most engineers don't have and most sales professionals don't have either.
Every project here is real code in a real repository — not a mockup. Click through to the source.
Persistent AI project assistant / multimodal wearable AI
A local-first AI assistant with application-owned context — not model conversation memory. FastAPI backend with deterministic context retrieval, structured Project/Activity/Checkpoint state, evidence provenance, and multimodal Investigation workflows that fuse screenshot evidence and spoken context into concise, glasses-ready guidance.
View repository →Reproducible ML workflow with Git + DVC
An MLOps demonstration project separating source code (Git) from data and model artifacts (DVC) so a teammate can clone the repo, run dvc pull, and reproduce an identical trained model and metrics — the reproducibility discipline that matters more than the model itself.
View repository →Supervised classification for telecom plan upsell
Classifies prepaid subscribers into Smart vs. Ultra plans from monthly usage behavior. Compared Logistic Regression, Decision Tree, and Random Forest with manual hyperparameter search; final model hit 79.8% test accuracy against a 75% target and a 69.4% majority-class baseline.
View repository →Imbalanced classification with upsampling
Predicts bank customer churn from 10K records, addressing ~80/20 class imbalance via class weighting and upsampling. Random Forest with upsampling was the best performer: F1 ~0.61 (exceeding a 0.59 threshold) and AUC-ROC ~0.86.
View repository →Statistical EDA on prepaid plan revenue
Exploratory and statistical analysis of the same Megaline subscriber base to determine which prepaid plan — Smart or Ultra — generates more revenue, pairing directly with the plan-recommendation classifier above.
View repository →Customer ordering behavior at scale
Exploratory analysis of the Instacart 2017 grocery dataset covering order timing, product/department popularity, and reorder behavior, with business-focused findings on cart composition patterns.
View repository →A decade in enterprise technology sales and solution engineering, across SaaS, cloud, and infrastructure — the customer-facing foundation this pivot is built on.
BILL.com · Houston, TX
Hewlett Packard Enterprise · Dallas–Fort Worth, TX
Sirius Computer Solutions · Dallas–Fort Worth, TX
Sirius Computer Solutions · Dallas–Fort Worth, TX
Fastest way to reach me is email. The form below sends straight to my inbox.