Kyle Rauch

Kyle Rauch

AI-augmented technical operator

Download PDF resume

Summary

AI-augmented technical operator. 4+ years L2 SaaS support depth at Clubessential plus 18 months financial services at Fidelity (FINRA Series 7/63), now shipping production AI tooling for the exact problems my support work surfaced — 16 public repos in 2026 spanning B2B support triage, CSM dashboards, calibrated ML + grounded LLM systems, and multi-provider AI facades. Direct AI to implement; own the architecture, eval, and shipping. Targeting Solutions Engineer / AI Implementation / AI-augmented CSM at AI-forward SaaS.

Professional Experience

Independent AI Engineering Portfolio

Feb 2026 – Present

Self-directedCincinnati, OH

  • Lead project: OfferPilotV2 — desktop-first, local-first job-search command center in Tauri v2 + React / TypeScript + Rust. OS-keychain integration, atomic-write contracts, file-based Codex bridge for AI-assisted generation, Vitest + Playwright + native Rust test stack.
  • Shipped 16 public repos in 2026 spanning B2B support triage automation, CSM dashboards, calibrated ML with grounded LLM layers, multi-provider AI facades (Anthropic / OpenAI / Gemini / Groq behind one structured-call interface), and agentic orchestration frameworks.
  • Every project ships with CI, an eval harness, and documentation showing prompt-version receipts.
  • Concurrent: BS Information Technology at Western Governors University (in progress).

Level 2 Website Support Specialist

Oct 2021 – Jan 2026

ClubessentialBlue Ash, OH

  • L2 / L3 escalation point across websites, reservation systems, CRM, and mobile apps for the 1,000+ private-club SaaS client base. Carried an average 90 tickets/week through Salesforce.
  • Partnered with engineering on root-cause analysis for recurring technical issues; drove long-term platform fixes rather than per-client workarounds across the website, reservation, and CRM stacks.
  • Hands-on HTML, CSS, and SQL daily: hand-edited HTML to fix client web pages and embedded tables; wrote CSS within the reservation system’s templating layer; built SQL queries against the CRM to extract form-submission data and run client-requested reports.
  • Client-facing platform optimization and staff training across onboarding and re-engagement cycles for new and existing clubs.

Customer Relationship Advocate

Apr 2019 – Sep 2020

Fidelity InvestmentsCovington, KY

  • Delivered client support for account maintenance, trade execution, and money movement in a high-compliance financial services environment.
  • Applied structured problem-solving to resolve complex financial inquiries while maintaining FINRA / SEC compliance.
  • Earned FINRA Series 7 and Series 63 licenses (now expired; reactivatable).

Logistics Account Executive

Oct 2017 – Mar 2019

Total Quality LogisticsCincinnati, OH

  • Managed client accounts and negotiated freight services in a high-volume outbound sales environment (75+ calls/day).
  • Built and held client relationships through daily outreach and structured pipeline management.

Selected Builder Work

/

OfferPilotV2

Tauri v2 · React · TypeScript · Rust · OpenAI Responses API

Desktop application owning the application-packet pipeline end-to-end: posting fetch with explicit consent, packet request creation, tailored .docx + cover .pdf generation, review-checklist gating, pipeline lanes with persistent state. Two generation paths (file-based Codex bridge + direct API with OS-keychain key storage). Vitest + Playwright + native Rust test stack.

/

diamond-edge

Python · LightGBM · Anthropic · isotonic regression

Gradient-boosted models on Statcast + odds; rationale generation constrained to top-k SHAP attributions (no free-form storytelling). Isotonic calibration reduced max deviation 14.3% → 5.6%, ECE 0.065 → 0.0004. Variance-collapse guardrail caught a silent "model shipped as market passthrough" bug (CLV −1.045%). Built in 3 weeks with dual-agent dev workflow.

/

support-triage

Python · FastAPI · React · FAISS · Anthropic

Classifies tickets, retrieves KB context via embeddings, drafts citation-grounded replies, surfaces top-3 macros — the AI version of the work I was doing at Clubessential. Latest eval: 97.1% strict faithfulness, 95.8% recall@3, 95% category accuracy, 100% within-1-tier priority. Single-operator architecture; CI + reproducible eval runbook.

/

csm-account-pulse

Python · Streamlit · Anthropic

Aggregates CRM/usage/ticket data, computes health scores, generates 3-bullet "focus this week" briefings per account with every claim citing a real fixture field. Shipped through 5 documented prompt iterations (v1 → v5) with eval receipts; precomputing day-counts in the LLM payload eliminated v3 hallucinations. Bring-your-own-CRM via four-CSV CsvDataSource. CI.

Sixteen public repos at github.com/kgr1115 — also SiftRobust (AI inbox triage), QuarterlineV2 (CRE research platform), understudy (agentic job-search orchestration), and project scaffolding tooling.

Education

Western Governors University

BS, Information Technology

In progress

Moeller High School

High School Diploma