~/cjfravel.dev
scala · utf-8 · ● live
cj@fravel:~ $ whoami

$ CJ Fravel

trait SoftwareEngineer extends Engineer with Pragmatist
problems.flatMap(decompose).map(build).filter(_.ships)

Software engineer who likes hard problems. Build data platform foundations other engineers ship on — declarative data contracts that power the platform's core data processing on Spark, turning raw data into canonical, governed datasets — with an evaluation-driven workflow around coding agents. Web/backend background from a previous life. Big proponent of agile, CI/CD, testing, and clean code.

open to interesting problems Greater Seattle Area data · analytics · platforms
~/experience.log
tail -f · 3 entries

// experience — git log --oneline

Microsoft / Senior Software Engineer
Aug 2021 — Present

Platform ownership: proposed and own a declarative data-contract system that is the platform's core data-processing stage — it replaced the entire legacy U-SQL processing layer in the migration to Spark on Synapse — and teams use it to produce canonical, governed datasets (schemas, validation, governance, delivery SLAs) self-served across Partner, Seller, and Royalties orgs without modifying platform code.

Data-integrity quality gates: replaced manual contract-review checks with automated validation built into the pipeline, removing humans from the critical path while raising the quality bar across Partner, Seller, and Royalties orgs.

Reporting & metrics: build the ADX (Kusto) dashboards that track SLA, performance, and COGS — the primary way platform delivery health and cost are reported.

Unified cross-team reporting: built Spark observers that aggregate cross-team metrics into a unified reporting layer, and lead cross-team coordination on the next-generation platform for the incentives space.

Resilient orchestration: Azure Data Factory and Synapse pipelines, plus custom Azure Functions orchestration for cross-region redundancy — pipelines fail over to a backup region when the primary is down and resume on restore.

Developer experience: drive faster local Spark workflows, self-service onboarding for new entities and teams, and reliability tooling (SLA / quarantine / alerting) so daily work is predictable.

Production at scale: the platform processes 2+ TB/month representing $110B in revenue; the Spark/Synapse rebuild delivered ~8× E2E speedup at 1/5 the cost.

Also helped move an existing big-data platform from the commercial cloud to a secure federal environment.

Applied Information Sciences / Software Engineer
May 2019 — Aug 2021

Lead developer on a Microsoft-funded POC for the Air Force Test Center, demonstrating how Azure Storage, Data Lake, Data Factory, and Jupyter notebooks streamline analytics across traditionally siloed test-wing data.

Technical team lead for a Data Integration Platform: secure cloud ingest → automated pipelines → Power BI dashboards delivering self-serve metrics to stakeholders across previously siloed data.

Pixel LLC / Developer
May 2017 — May 2019

Employee #1 — founding-team experience at a small startup: helped build the team and culture and did the late-night 0→1 work, alongside customer-facing web and mobile consulting shipped directly to end users.

~/open-source/
3 repos · git: clean

// open source — things I build in the open

cjfravel-dev / ariadne Scala

A Spark utility for generating indexes in a data lake to reduce the amount of data included when joining across massive datasets. Useful when data is poorly partitioned (or not at all) and cannot be moved, or when duplication is too costly.

cjfravel-dev / ChromaMark JavaScript

A lean rich-markup format for agent-to-human communication — a strict superset of Markdown that adds colored blocks, status pills, and collapsible sections at a fraction of the token cost of equivalent HTML. Ships a JavaScript renderer (npm/CDN), a byte-identical Python port with Jupyter support, a CLI, and a VS Code extension.

→ view this profile rendered in ChromaMark — dogfooded, live via the CDN

cjfravel-dev / nomos Scala

A JSON templating engine for Scala. Define a template once and nomos generates matching case classes that validate, serialize, and deserialize JSON — one source of truth for your model and its validation, with a small, dependency-free runtime.

~/ai_workflow.md
cat · markdown

// working with AI — evaluation-driven, expertise-led

I'm an AI-first engineer with an evaluation-driven workflow around coding agents. Direction, architecture, and judgment calls are mine; agents execute the well-scoped pieces. The leverage comes from knowing how things work.

  • Verification-first & safety-minded: heavy automated testing and validation against real production numbers, so what ships behaves as claimed — catching confidently-wrong output before it reaches production
  • Evaluation-driven workflow: scoped task decomposition, prompt iteration, sandbox validation, diff-based human-in-the-loop review, and production guardrails on every change agents land
  • Agent loops & tool design: shape the agent loop and tool-calling surface so the agent has the right primitives for the job; tighten the loop where it drifts
  • Autonomous workflows: design automation and scaffolding that let agents run hands-off on appropriate work, protecting reliability and developer productivity
  • Design partner & research: pressure-test architecture and surface tradeoffs before committing; explore unfamiliar tech with grounded citations
  • Benchmarking & iteration: treat each task as an informal benchmark; measure outcomes, iterate on prompts, tools, and evaluation frameworks

// guardrails: review every diff, keep secrets out of prompts, sandbox before prod, and stay fluent enough to catch when the agent is confidently wrong.

~/education.yaml
yaml · read-only

// education & certifications

Wright State University / Bachelor's in Computer Science
Dayton, OH

Learned to effectively and efficiently develop software, analyze programs, and apply familiarity with real-world packages and tooling.

CompTIA Security+ · Jun 2019 — Jun 2022
Microsoft Certified: Azure Data Engineer Associate · Mar 2021 — Mar 2023