Alex Egger
Menu
ProofExperienceCasesFocusStackReposEducation
Contact
Mid-level AI / full-stackNext.js · Python · LangGraph · AWSRemote

I build production RAG workflows and multi-tenant AI products. I focus on reliable architecture, evaluation-driven delivery, and clean CI/CD.

Alex EggerMid-level AI and full-stack engineer | Next.js, Python, LangGraph, AWS

  • Plan review on jurisdiction-specific RAG dropped from hours to minutes per document.
  • A logistics invoicing workflow moved from about five days to same-day processing.
  • Pytest and Playwright suites gate every deploy behind a green CI pipeline.

I have spent the last three years building production RAG workflows, agent systems, and multi-tenant SaaS on AWS. I work across the stack: Next.js on the frontend, Python and TypeScript services, PostgreSQL with pgvector, auth and billing, and CI/CD with GitHub Actions. I started in quantitative finance, so I naturally focus on measurable outcomes and solid evaluation practices.

I am looking for mid-level AI or full-stack roles with meaningful AI systems work. Remote in the US is ideal.

EmailLinkedInGitHubResume (source)

Business impact

Case studies and outcomes

Each case maps to a concrete line on my resume, with context on where the result came from.

Plan review: hours to minutes

Built production LangGraph RAG over blueprints, specifications, and jurisdiction-specific building codes. Typical plan review dropped from several hours to minutes per document.

Scope and definition

Pre-seed startup (AEC / construction tech), founding engineer, document review cycle vs. prior manual workflow (Oct 2025 to present).

Invoices: five days to same-day

Built an invoicing pipeline using OCR plus rule-based validation. It reduced one logistics client's payment cycle from about five days to same-day.

Scope and definition

Independent consultant, AI automation engineer, SMB finance ops (Feb to Sept 2025).

Green CI before every deploy

Wrote integration and regression suites in pytest and Playwright, then used them to gate every deploy.

Scope and definition

Pre-seed startup (AEC / construction tech), founding engineer, release process (Oct 2025 to present).

Stronger backtests (options)

Refined options-volatility signal generation and improved out-of-sample Sharpe and CAGR in backtests, alongside a Nuxt plus Python analytics platform.

Scope and definition

APX Financial Analytics · full stack, financial analytics · systematic strategies; backtests only, not live trading advice (2024).

Less manual CRM work

Automated lead intake and CRM synchronization in n8n, removing a substantial block of weekly manual data entry from the client's operations team.

Scope and definition

Independent consultant, AI automation engineer, sales and ops workflow (Feb to Sept 2025).

Experience

Professional experience

Pre-seed product engineering, consulting delivery, financial analytics, and quant research. Newest first.

Pre-seed startup (AEC / construction tech)

Founding Engineer · Overland Park, KS

Oct 2025 - Present

  • Founding engineer leading development of a multi-tenant SaaS for blueprint analysis, code compliance checks, and construction document generation, deployed with municipalities, AEC firms, and private contractors.
  • Built production RAG pipelines on LangGraph over blueprints, specifications, and jurisdiction-specific building codes; typical plan review dropped from several hours to minutes per document.
  • Implemented agent reliability patterns: Pydantic schema validation, exponential-backoff retries, deterministic tool routing; traces and failure modes monitored through LangSmith and CloudWatch.
  • Built the stack end-to-end: Next.js frontend, Python and TypeScript services, PostgreSQL with pgvector, tenant isolation, auth, billing, and CI/CD via GitHub Actions on AWS ECS.
  • Wrote integration and regression test suites with pytest and Playwright, gating every deploy behind a green pipeline.
  • Hired and onboarded the second engineer; set code review, sprint cadence, and testing standards for the team.

Independent Consultant

AI Automation Engineer · Kansas City, MO

Feb 2025 - Sept 2025

  • Built an invoicing pipeline combining OCR extraction with rule-based validation, reducing a logistics client's payment cycle from roughly five days to same-day.
  • Automated lead intake and CRM synchronization in n8n, removing a substantial block of weekly manual data entry from the client's operations team.
  • Instrumented multi-agent outputs with structured validation, retry logic, and a custom eval harness run on every prompt or model change before shipping.
  • Delivered technical audits and AI implementation roadmaps for SMB clients across hospitality, logistics, construction, finance, and wholesale.

APX Financial Analytics

Full Stack Engineer, Financial Analytics · Kansas City, MO

Jan 2024 - Dec 2024

  • Built a full-stack analytics platform (Nuxt + Python) for internal trading teams, integrating IBKR, ORATS, and CBOE market-data feeds over REST.
  • Refined signal generation on options volatility strategies, improving out-of-sample Sharpe and CAGR in backtests across equity index markets.
  • Developed ML models for volatility forecasting and options pricing to support systematic strategy research.
  • Automated analytics and reporting pipelines, removing a recurring weekly manual reporting burden for the research desk.

Deep dives

Start here on GitHub

Three repos that are easy to discuss in interviews. Start with the README, then look at tests and CI.

LLM gateway

README

OpenAI-compatible edge with policies, retries, circuit breakers, idempotency hooks, and streaming. Start with the README.

llm-traffic-controller

RAG workbench

README

Comparable retrieval paths, mock-friendly runs, and a regression-friendly harness for LangGraph RAG experiments.

langgraph-rag-workbench

Agent traces

README

Supervisor squad with normalized trace export and a small console. This is how I debug multi-step agents.

supervisor-squad-trace

Focus areas

How I group the work

Three lanes: edge reliability, spend control, and RAG or agent runtime. Each card links to repos you can clone and run.

Gateway, policy & reliability

OpenAI-compatible gateways and policy graphs: retries, circuit breakers, idempotency, streaming backpressure, and explicit pre/post checks so LLM behavior stays predictable under real traffic. Reference repos mirror patterns I use in production (YAML policy, schema validation, mock-friendly CI paths).

FinOps and tenancy

I treat cost controls as part of product quality. This lane covers metering, budgets, and tenant-aware usage tracking, plus operator tooling teams can actually use.

Stack

Skills

Grouped the same way as my resume. Easy to scan, and easy to verify in the repos.

Languages

PythonTypeScript/JavaScriptSQL (PostgreSQL, MySQL)

AI/ML & LLMs

LangChainLangGraphRAGMulti-agent orchestrationMCPMultimodal pipelinesOpenAIAnthropicHugging Face

LLM observability & evals

LangSmithLangfuseRagasBraintrustCustom eval harnesses

Data & vector stores

PostgreSQLMySQLpgvectorPineconeChromaQdrant

Cloud & infra

AWS (EC2, S3, ECS)DockerGitHub ActionsLinux

Testing

pytestPlaywrightIntegration and regression test design

Frontend

ReactSvelteNext.jsNuxtTailwind CSS

Automation & integrations

REST APIsn8nOCR pipelines

Portfolio

Selected repositories

A focused set of projects first, then additional explorations if you want to go deeper.

LLM traffic controller

llm-traffic-controller

GatewayPostgresPrometheus

OpenAI-compatible gateway: YAML policies, retries, circuit breakers, dual upstreams, idempotency hooks, Postgres metering, Prometheus.

Open on GitHub

LangGraph RAG workbench

langgraph-rag-workbench

LangGraphRAGEvals

RAG + supervisor patterns; NDJSON streaming; RAG_WORKBENCH_MOCK=1 runs the UI without OpenAI spend.

Open on GitHub

Agent runtime kernel

agent-runtime-kernel

RuntimeLangGraphMetrics

Multi-tenant job runtime for LangGraph: scheduler, cancel, deadlines, sandboxed tools, /metrics.

Open on GitHub

Policy graph guard

policy-graph-guard

GovernanceLangGraphOpenAPI

LangGraph policy DAG around chat: PII/topic/tools/injection pre, LLM, JSON schema + secret post; MOCK + strict; OpenAPI.

Open on GitHub

Workflow contract bridge

workflow-contract-bridge

WebhooksOutboxSvelteKit

Versioned JSON Schema webhook ingress; dead letters + replay; transactional outbox; FastAPI + SvelteKit admin.

Open on GitHub

LLM cost guard

llm-cost-guard

FinOpsNext.jsPostgres

Multi-tenant metering + daily USD budgets; Postgres, guarded OpenAI client, Next.js ops UI, LangGraph demo; prompts off by default.

Open on GitHub
View additional explorations

Role Fit Coach

rolefit-coach

SvelteKitLangGraphPostgres

SvelteKit BFF, JWT httpOnly cookie, SSE LangGraph coach, Postgres.

Open on GitHub

Incident copilot

incident-copilot

HITLLangGraphCI

HITL incident workflow with interrupt and resume paths, plus mock mode for CI.

Open on GitHub

Horizon task decomposer

horizon-task-decomposer

DAGHITLSafety

Plan → validate → execute → verify over a JSON task DAG; HMAC approvals; NDJSON; HORIZON_MOCK offline path.

Open on GitHub

Supervisor squad trace

supervisor-squad-trace

LangGraphTracingOTLP

Supervisor squad with trace export (NDJSON + optional OTLP) and Next.js console.

Open on GitHub

Invoice reconcile pipeline

invoice-reconcile-pipeline

FastAPIOCRPostgres

AP demo: PDF + CSV PO → arq worker, rapidfuzz explainable matches, exception queue, Postgres audit; FastAPI + Next.js + Compose.

Open on GitHub

Scout (Rust TUI)

scout

RustTUIAgents

Rust TUI agent with allowlisted HTTP and an optional LLM planner.

Open on GitHub

Education and highlights

Education, certifications, and competitions

Coursework and certifications match the resume source. Competitions listed here are also in the resume.

Education

Park University - Business and Finance coursework (2022-2025), Parkville, MO

Certifications

  • DeepLearning.AI - Deep Learning Specialization (Coursera)
  • Stanford Online - Data Science and Machine Learning (Coursera)
  • HarvardX - Machine Learning (edX)

Optiver Quantitative Finance Competition

Python, PyTorch, scikit-learn - Finalist

  • Options volatility forecasting with ensemble ML methods and engineered features on high-frequency order-book data; selected as a finalist.

Kaggle Cardiac Risk Detection

Python, TensorFlow, Deep Learning - Top submission

  • Deep learning model for cardiac risk detection from patient signal data; preprocessing and augmentation to improve recall on highly imbalanced medical datasets.