AI & Backend Engineer — Arlington, VA
Jose Alvarado Alvarenga
Product-minded engineer building event-driven systems and developer tools at Capital One — and shipping AI-native products on my own time. I ground LLM features in structured data, strict contracts, evals, and production observability.
4+
years building
AWS
certified architect
v1.0
live on the App Store
PROMOTER / Now
Currently in the lab.
The active build — a 12-week public sprint at the intersection of the biology degree and the AI engineering.
ACTIVE BUILD — SUMMER 2026
A citation-grounding evaluation harness for oncology claims.
LLMs cite sources that don't fully support their claims — a documented, still-open failure mode, and in medicine a dangerous one. I'm building an eval harness that decomposes model answers into atomic claims, retrieves the cited sources, and scores whether each claim is actually supported.
This is where the biotechnology degree stops being a fun fact: judging whether a source supports an oncology claim requires knowing the biology. That's the part no generic pipeline can do.
IN THE HARNESS
Claim decomposition & retrieval over PubMed oncology literature
Expert-labeled ground truth — built by hand, small by design
LLM-as-judge alignment, failure taxonomies, regression suites
Fine-tuned baselines vs. prompted frontier models
Building in public — follow along on GitHub
EXON 01 / Selected work
Things I've shipped.
AI-native products with real users and real constraints: structured outputs, deterministic validators, replayable evals, and MCP surfaces — not demos.
RepNotes AI
An AI-native iOS workout tracker that turns free-form workout notes into structured logs, progress analytics, insights, and next-workout plans.
PROBLEM
You're between sets with 60 seconds on the clock. Every workout app wants you to tap through menus and dropdowns to log what you just did. Notes apps are fast, but they leave you with a pile of unstructured text that can't track progress.
APPROACH
Log your workout the way you'd text a friend. Parser v2, an AI-assisted understanding pipeline, normalizes messy natural-language notes into exercises, sets, reps, weights, and cardio — and Plan AI turns that history into grounded insights and next-workout plans.
HIGHLIGHTS
Parser v2: normalizes messy natural-language notes into exercises, sets, reps, weights, cardio, and training facts — with correction replay
Plan AI: facts-grounded coaching built on LLM planner/narrator stages, strict JSON contracts, and deterministic evidence validators
Rejects unsupported, stale, or fabricated recommendations before they ever reach the user
Replayable AI evals and privacy-safe diagnostics covering raw-text leakage, stale data, unsupported claims, and evidence drift
Whoop MCP Server
A remote MCP server that securely connects Whoop biometric data to Claude for personalized endurance coaching.
PROBLEM
You're training for a race with a Whoop on your wrist collecting HRV, sleep, recovery, and strain around the clock. All that data — and the decision of whether to train hard or rest still comes down to a gut feeling.
APPROACH
Connect your Whoop to Claude through the Model Context Protocol. Instead of staring at recovery scores and guessing, you ask Claude what to do today and it answers from your actual HRV, sleep, and strain data.
HIGHLIGHTS
Remote MCP server securely bridging Whoop biometric APIs and Claude via OAuth 2.0
7 tools for pulling and analyzing recovery, HRV, and sleep metrics
Computes acute-to-chronic workload ratio (ACWR) and cumulative sleep debt
Generates personalized daily Ironman 70.3 coaching recommendations
EXON 02 / Experience
Four years of systems that stay up.
Jan 2024 — Present
Senior Associate Software Engineer
Capital One · McLean, VA
Engineered a custom CLI that gives Claude and other AI agents on-demand access to decentralized documentation — faster information retrieval for engineers and less onboarding friction.
Architected an event-driven fan-out system (Lambda, SNS, SQS, DynamoDB) and AWS Glue ETL pipelines orchestrating multi-channel marketing engagement, customer segmentation, and offer fulfillment.
Helped build and launch an internal AI tooling marketplace — a centralized hub where 50+ developers discover and share custom Claude Code skills and agents.
Led platform migrations (RDS to DynamoDB, EC2 to Fargate) coordinating across 3 teams, delivering multi-region reliability and 15% performance gains.
Management-nominated member of the AI Trailblazer program, defining engineering AI-adoption practices and presenting findings to senior leadership.
Mentored 5 new associates in the Technology Development Program and drove GenAI-powered observability improvements that cut high-severity incidents 10%.
Aug 2022 — Jan 2024
Associate Software Engineer
Capital One · Richmond, VA
Championed the modernization of 3 legacy APIs from Java 8 to 17 — navigating technical debt and stakeholder priorities to deliver a 30% latency reduction and improved system stability.
EDUCATION
B.S. Biotechnology, Minor in Computer Science
James Madison University · 2018 — 2022
CERTIFICATION
AWS Certified Solutions Architect — Associate
Amazon Web Services · 2023
EXON 03 / About
Models write prose. Systems own the facts.
I'm a product-minded AI and backend engineer — 4+ years at Capital One building event-driven systems, data pipelines, and developer tools, and nights and weekends shipping RepNotes AI, now live on the App Store.
The through-line in everything I build is grounding LLM features in structure: strict JSON contracts, deterministic validators, replayable evals, and production observability. At work that looks like a CLI that gives AI agents on-demand access to decentralized docs and a marketplace where 50+ developers share Claude Code skills. At home it looks like an AI coaching pipeline that rejects any recommendation it can't back with evidence.
I studied biotech, which sounds like a left turn, but really I've just always been obsessed with how systems work and how to make them perform better. These days the test subject is me.
Languages
Python
TypeScript
Swift
Java
SQL
AI Engineering
OpenAI Structured Outputs
Claude / Claude Code
MCP
LLM Evals
Agent Tooling
AWS
Lambda
SNS / SQS
DynamoDB
Glue
Fargate
Multi-Region
Backend & Data
Event-Driven Systems
Microservices
ETL Pipelines
PostgreSQL / Supabase
Splunk
Product
SwiftUI
RevenueCat
Sentry
OAuth 2.0
INTRON / OFF THE CLOCK
Ironman 70.3
2027 — race TBD
Richmond Half Marathon
after that
I'm training with my own tools in the loop — the Whoop MCP server reads each morning's recovery and Claude decides whether I swim, bike, run, or rest. Dogfooding, but with a heart-rate monitor.
3′ UTR / Contact
Let's build something grounded.
Open to new opportunities and collaborations — or just talking shop about AI products, MCP, and endurance training.
CTG·ATC·TTC·GAG
© 2026 Jose Alvarado Alvarenga