Andrew Bush

DevOps Engineer

SeniorAI/ML Engineer

6/10Containerization & Orchestration

7repositories checked by code analysis

3projects

Overview

Platform-focused SRE (senior-level) with a strength in building production-minded, observable services and secure AWS infrastructure. The strongest proven skill is designing resilient services and supporting infrastructure, evidenced by the Go checkout service with idempotency, retry/circuit logic and tests (grocery-checkout-reliability-lab/internal/checkout/service.go and service_test.go) and the Terraform AWS designs (aws-health-project/infra/main.tf). Public code does not show CI/CD pipelines, remote Terraform state and locking, advanced alert routing or chaos/rehearsal exercises.

Timeline

Cloud Engineer (Part-Time Volunteer) · Middle

Elite Web Services · Part-Time

Partially remote

Helps build an AWS-based cloud optimization platform in a learning environment, including application features and operational runbooks. Works on backend/serverless components and integrates AWS services for authentication, monitoring, and alerting. Collaborates on API and frontend integration using TypeScript/JavaScript and React.

JavaScript

TypeScript

React.js

AWS Lambda

API Gateway

CloudFormation

Amazon CloudWatch

Media AI Trainer (Part-Time Contract) · Middle

LinkedIn · Contractor

Partially remote

Develops realistic media and video tasks to help train and evaluate AI systems in line with professional production workflows. Writes detailed prompts, acceptance criteria, and reference outputs, then evaluates AI-generated media for quality, accuracy, consistency, and creative-brief alignment. Provides structured feedback on model limitations and proposes validation checks to improve evaluation reliability.

Savannah College of Art and Design

Bachelor's Degree · Film & Television

2002–2006 Atlanta, Georgia

Transcode Operator · Middle

Deluxe Digital Studios · Full-Time

North Hollywood In office

Operated enterprise digital media transcoding workflows for major studio content to prepare assets for downstream post-production and distribution. Performed media ingest, format conversion, quality control checks, and delivery verification for high-volume production pipelines. Ensured processed outputs met requirements for subsequent mastering and distribution steps.

Media Manager / Post-Production Systems (Freelance / Part-Time, Remote) · Lead

OvationTV · Freelance

Los Angeles Partially remote

Delivered remote support for enterprise media workflows across shared SAN/NAS environments used in broadcast and post-production. Coordinated ingest, transcoding, delivery, archival, and retrieval activities while assisting editors and production teams. Handled storage management and troubleshooting for large-scale shared media access.

Media Manager / Post-Production Systems · Lead

OvationTV · Full-Time

Los Angeles In office

Managed enterprise media workflows on SAN/NAS shared storage for broadcast production, editorial, and content delivery. Administered and troubleshot approximately 100 TB of shared media storage, including file organization, capacity management, and operational support. Supported ingest, transcoding, delivery, archive restoration, and shared-storage access using media asset management tools and LTO-based archiving.

Media Manager / Studio Photographer · Lead

FracturedFX · Full-Time

Monrovia In office

Built and maintained the studio media department infrastructure from the ground up, including ingest workflows, naming conventions, storage architecture, and delivery pipelines. Designed and operated a long-term file-based archival system with searchable metadata and secure backup protocols. Managed version control, quality checks, and delivery of thousands of production assets across concurrent projects.

3D Capture Technician · Middle

Global Objects · Full-Time

North Hollywood In office

Produced high-resolution 3D and drone-photogrammetry assets for VFX and virtual production deliverables. Operated laser scanning and LiDAR-based capture equipment to acquire detailed spatial data. Processed and optimized 3D outputs using specialized capture-to-model workflows and supported downstream VFX rendering coordination.

Agisoft Metashape

Unreal Engine

Blender

AV Technician / Media Systems Support · Middle

Exhibition Hub · Full-Time

Los Angeles In office

Deployed and supported immersive AV and media systems across multi-room exhibition installations. Designed end-to-end AV and network infrastructure, including IP routing, VLAN configuration, and signal distribution for operational handoff. Troubleshot complex AV and network issues under strict pre-show deadlines across multi-vendor environments.

TCP/IP

VLAN

Projects

AWS AI Cloud Operations Copilot

Built an AWS AI agent for automated cloud incident investigation using Bedrock and LangGraph. Integrated operational data sources such as CloudWatch and storage-based runbooks, while applying least-privilege IAM for secure troubleshooting. Deployed the infrastructure with Terraform and provided a Streamlit UI along with automated test coverage.

SchoolVault (Serverless Media Archive)

Architected and deployed a serverless cloud media asset management platform for K-12 institutions. Built a searchable, access-controlled archive that replaces scattered local storage with centralized cloud storage and delivery. Integrated AI-assisted tagging and search capabilities and ensured the solution was hardened to pass a third-party penetration test remediations.

Highly Available AWS Web Infrastructure

Designed a highly available AWS environment using multiple Availability Zones for resilience under traffic demand. Implemented core networking and scaling components, including VPC subnets, load balancing, and automated scaling behavior. Added data durability and monitoring by configuring Multi-AZ database usage, asset hosting, and alerting for operational visibility.

Technical skills

Languages

Python

C++

JavaScript

TypeScript

DevOps

AWS

Amazon S3

Amazon ECS

Amazon CloudWatch

AWS Lambda

Kubernetes

Amazon EC2

TCP/IP

API Gateway

Amazon EKS

GitHub

IaC

Containers

Docker

Git

Linux

Terraform

VMWare

Windows

VLAN

CloudFormation

AI/ML

AWS Bedrock

Streamlit

AI Agents

LangGraph

LangChain

Copilot

Other

Boto3

Unreal Engine

SQLite

React.js

DynamoDB

Blender

Agisoft Metashape

Adobe Photoshop

DHCP

DNS

IAM

Adobe Premiere Pro

Adobe After Effects

Least Privilege

Senior AI/ML Engineer Confidence: High LLM Engineer

LLM engineer at a senior level specializing in safe, evidence-gated agent workflows for cloud-operations use cases. The strongest proven skill is building constrained, test-covered LLM orchestration and safety gates, demonstrated by app/agent/workflow.py and the extensive offline tests in app/tests/test_evaluation_prompts.py. The public codebase does not show custom model training, GPU/quantization optimization, or mature production MLOps pipelines such as model versioning, drift monitoring, or large-scale deployment automation.

Model Architecture & Training

1/10

How well models are designed and trained

No custom model training or architecture; engineering focuses on safe LLM orchestration, prompt/system prompt design, and bounded model calls.

Evidence

aws-cloud-operations-copilot/app/agent/workflow.py: generate_report calling bedrock_client.converse with SYSTEM_PROMPT and inferenceConfig

aws-cloud-operations-copilot/app/tests/test_evaluation_prompts.py: RecordingBedrockClient.converse capturing modelId and usage

Data Pipeline & Feature Engineering

3/10

How data is prepared for models

Practical, well-structured ETL for a training lab: reads Excel XLSX internals, validates headers, converts types and populates a local SQLite DB and CSVs.

Evidence

water-prep/sql/build_sqlite.py: read_raw_billing/read_cost_centers, typed_rows and connection.executemany to populate cloud_billing

water-prep/sql/run_query.py: read-only SQLite prompt and query/display logic

Experimentation & Evaluation

4/10

How results are measured and tested

Clear unit tests and an offline evaluation harness that validate prompt boundaries, evidence gating, and workflow behavior; test coverage exercises multiple important code paths and safety checks.

Evidence

aws-cloud-operations-copilot/app/tests/test_evaluation_prompts.py: offline evaluation cases and assertions about model invocation and scoping

aws-cloud-operations-copilot/app/tests/test_agent_workflow.py: tests for evidence gating, bypassing model calls, and correct usage limits

MLOps & Deployment

3/10

How models are shipped to production

Deployment-oriented glue exists (Streamlit UI, boto3 Bedrock client, a Lambda entrypoint and Terraform docs) but no end-to-end MLOps (no model versioning, drift monitoring or production orchestration evidence).

Evidence

aws-cloud-operations-copilot/streamlit_app.py: build_workflow_from_environment building a Bedrock client and wiring the workflow

aws-cloud-operations-copilot/app/lambda_src/app.py: lambda_handler (runtime entrypoint)

Computational Efficiency

2/10

How efficiently computing resources are used

Some efficiency/cost controls at the orchestration level (token caps, single bounded call) but no GPU/quantization/distributed optimization work.

Evidence

aws-cloud-operations-copilot/app/agent/workflow.py: MAX_OUTPUT_TOKENS and limiting when model is called

aws-cloud-operations-copilot/streamlit_app.py: model_usage_label and UI cost guard

Research Depth & Innovation

2/10

Depth of research and new ideas

No novel research or SOTA implementations; shows thoughtful safety and guardrail design for LLM usage and tool scoping but not original ML algorithm work.

Evidence

aws-cloud-operations-copilot/app/agent/workflow.py: SYSTEM_PROMPT enforcing evidence-only reasoning and explicit headings

aws-cloud-operations-copilot/app/diagnostic_tools/tools.py: redact_log_message and strict allow-listing in DiagnosticTools._require_allowed_service

Expertise

AI Agents & Agentic Workflows· Senior

MLOps & Model Lifecycle· Middle

Industries

Artificial Intelligence· Middle

Information Technology· Middle

Software· Middle

Technologies

Python· Senior

C++

Copilot

LangGraph

LangChain

Terraform

DynamoDB

VMWare

CloudFormation· since 2026

AI Agents

AWS Bedrock

Git

SQLite

Docker

Kubernetes

Streamlit

Boto3

Amazon EKS

AWS Lambda· since 2026

Amazon EC2

GitHub

Amazon S3

IAM

Amazon CloudWatch· since 2026

API Gateway· since 2026

Linux

Windows

TCP/IP· since 2025

DNS

DHCP

VLAN· since 2025

AWS· mentioned only

Copilot· mentioned only

Recommendations

Develop safe, explainable incident-response agents that integrate constrained read-only system tooling and human review (e.g., evidence-gated Bedrock agents).
Build developer-facing cloud-operations labs and FinOps analysis tooling that combine deterministic data pipelines with reproducible evaluation harnesses.
Implement production MLOps features around model versioning, observability, and cost/usage tracking to evolve the current proof-of-concept into a deployable service.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

water-prep aws-cloud-operations-copilot kubernetes-application-delivery-platform

Senior DevOps Engineer Confidence: High Platform SRE

CI/CD Pipelines

2/10

Automated build and deploy

Minimal CI/CD automation present; only container build artifact and run scripts are visible but no reusable workflows, pipeline gates, or artifact signing are present.

Evidence

grocery-checkout-reliability-lab/Dockerfile

Infrastructure as Code

5/10

Managing servers with code

Reasoned Terraform designs for multiple AWS services with security and lifecycle settings are present, but no remote state backend, locking, versioned modules, or infra tests are visible.

Evidence

aws-health-project/infra/main.tf

grocery-checkout-reliability-lab/infra/main.tf

aws-health-project/infra/.terraform.lock.hcl

Containerization & Orchestration

6/10

Working with containers

Good container and orchestration practices: multi-stage non-root Dockerfile, ECS task definitions with resource sizing, ALB health checks and autoscaling; application includes graceful shutdown handling. Missing Kubernetes-level artifacts like PodDisruptionBudgets or tuned probes beyond basic health endpoints.

Evidence

grocery-checkout-reliability-lab/Dockerfile

grocery-checkout-reliability-lab/infra/main.tf: aws_ecs_task_definition / aws_lb_target_group (health_check)

grocery-checkout-reliability-lab/cmd/server/main.go

Observability & Monitoring

6/10

Watching system health

Observability is deliberately instrumented: application exposes /metrics and increments structured metrics, CloudWatch dashboard and alarms are defined, and log groups with retention are configured. There is no evidence of multi-tenant alert routing, burn-rate SLO rules, or advanced alert deduplication.

Evidence

grocery-checkout-reliability-lab/internal/telemetry/metrics.go

grocery-checkout-reliability-lab/infra/main.tf: aws_cloudwatch_metric_alarm and aws_cloudwatch_dashboard

Reliability & Incident Response

6/10

Keeping systems up

Concrete reliability practices are implemented: idempotency handling, retries with backoff, a simple circuit-breaker pattern, tests that simulate failures, and graceful shutdown for clean deployments. Deploy strategies are basic (ECS deployment min/max and autoscaling) but meaningful incident/runbook artifacts or chaos experiments are not present in scored code.

Evidence

grocery-checkout-reliability-lab/internal/checkout/service.go

grocery-checkout-reliability-lab/internal/checkout/service_test.go

grocery-checkout-reliability-lab/cmd/server/main.go

Cloud & Cost Optimization

5/10

Smart use of the cloud

Cost and cloud discipline are visible via a budget, autoscaling target tracking, ECR scan-on-push, and S3 lifecycle/retention rules; however there is no evidence of spot/eviction strategies, fine-grained rightsizing history, or workload-identity migrations.

Evidence

aws-health-project/infra/main.tf: aws_budgets_budget

grocery-checkout-reliability-lab/infra/main.tf: aws_appautoscaling_policy / aws_appautoscaling_target

aws-health-project/infra/main.tf: aws_s3_bucket_lifecycle_configuration

Expertise

Site Reliability Engineering· Senior

Cloud Platforms & Architecture· Middle

Observability & Monitoring· Middle

Industries

Commerce· Middle

Health Care· Middle

Technologies

Containers

IaC

Go· Senior

AWS

Amazon ECS

Observability· mentioned only

SRE· mentioned only

Terraform· mentioned only

Recommendations

Own and expand SRE-facing platform work such as service reliability patterns, alert-to-runbook automation, and on-call playbooks for production services.
Build reusable IaC modules and add remote state with locking and automated infra tests (terratest or conftest) to harden multi-environment deployments.
Implement a small CI/CD pipeline with promotion gates, artifact signing and rollback steps to turn the Docker/ECS reference into a reproducible delivery flow.
Add SLO-driven alerting and burn-rate protection rules with documented rollback criteria to reduce noisy alerts and improve incident response.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

aws-health-project aws-healthops grocery-checkout-reliability-lab

Middle Backend Developer Confidence: High API Engineer

Backend engineer (senior-level evidence) focused on safe, small-scope cloud tooling and backend orchestration with a strong emphasis on deterministic LLM gating and security. The strongest proven skill is designing a constrained, test-covered LLM-driven incident workflow with concrete implementations in app/agent/workflow.py and app/diagnostic_tools/tools.py. Public code does not show production-grade scaling artifacts like distributed retries, caching/invalidation strategies, migration histories, or multi-service contracts for a large microservice landscape.

API Design

5/10

How well APIs are designed

Clear API contracts and deterministic model-invocation boundaries with input validation and protocol types; lacks documented versioning, idempotency keys, or pagination schemes.

Evidence

aws-cloud-operations-copilot/app/agent/workflow.py: DiagnosticToolsProtocol and build_incident_graph define the tool/workflow contract and single bounded model call

aws-cloud-operations-copilot/streamlit_app.py: get_runbook_bucket_from_environment validates required runtime settings before use

aws-cloud-operations-copilot/app/tests/test_agent_workflow.py: tests asserting model invocation parameters and scope

Data Layer & Database

4/10

Working with databases

Practical, well-typed local data handling and schema creation for exercises; no migration history or multi-stage schema evolution artifacts and only simple indexing choices.

Evidence

water-prep/sql/build_sqlite.py: creates cloud_billing and cost_centers tables, types rows, and creates indexes

water-prep/sql/run_query.py: read-only SQL prompt with explicit allowed prefixes and display limits

Scalability & Performance

3/10

Handling load and speed

Some attention to bounded queries and cost-control (token cap), but no evidence of caching strategies, queue-based decoupling, load testing, or connection-pooling work.

Evidence

aws-cloud-operations-copilot/app/diagnostic_tools/tools.py: get_recent_error_logs enforces time windows and max_events limits

aws-cloud-operations-copilot/streamlit_app.py: model_usage_label and invocation gating show cost guard awareness

System Architecture

5/10

Overall system structure

Deliberate decomposition into small modules (diagnostic tools, runbook retriever, workflow) and a StateGraph-based orchestration for a safe LLM workflow; scope is intentionally narrow and single-service-focused.

Evidence

aws-cloud-operations-copilot/app/agent/workflow.py: StateGraph with collect/retrieve/assess/generate nodes and clear routing logic

aws-cloud-operations-copilot/app/diagnostic_tools/tools.py: DiagnosticTools class encapsulates AWS interactions and contract enforcement

aws-cloud-operations-copilot/streamlit_app.py: build_workflow_from_environment composes modules only after validated inputs

Security & Auth

6/10

Protecting data and access

Strong security-aware choices for the project scope: allow-listing of resources, redact-before-exposing log text, configuration gates and tests guarding secret exposure; token and evidence gating for LLM calls.

Evidence

aws-cloud-operations-copilot/app/diagnostic_tools/tools.py: _require_allowed_service and redact_log_message enforce allow-list and secret redaction

aws-cloud-operations-copilot/app/tests/test_evaluation_prompts.py: tests assert SYSTEM_PROMPT prohibits secret exposure and automatic changes

aws-cloud-operations-copilot/streamlit_app.py: get_runbook_bucket_from_environment raises ConfigurationError if required env var missing

Reliability & Observability

5/10

Stability and monitoring

Good defensive error handling, deterministic short-circuiting (zero-token responses) and unit tests for recovery cases; limited evidence of retry/backoff, circuit breakers, distributed tracing, or operational runbooks for production scale.

Evidence

aws-cloud-operations-copilot/app/agent/workflow.py: logic that bypasses model calls for healthy or recovering evidence, producing deterministic low-cost responses

aws-cloud-operations-copilot/app/diagnostic_tools/tools.py: parameter bounds checks and explicit error types

aws-cloud-operations-copilot/streamlit_app.py: exception handling for AWS errors and ConfigurationError paths

Expertise

Backend AI & LLM· Middle

Python· Middle

System Architecture· Middle

Industries

Information Technology· Senior

Financial Services· Middle

Recommendations

Build small, production-ready LLM-enabled runbooks and incident responders that add retries, backoff with jitter, and structured metrics/tracing (use the existing agent/workflow.py as the safety-first base).
Develop FinOps ETL and analytics pipelines from the water-prep workbook into a repeatable data pipeline with migration history and schema evolution tracking.
Harden observability and resilience for the copilot: add retry/backoff, rate limiting, structured logs with correlation ids, and automated alerts tied to the zero-token gate behavior.
Expand API contracts with explicit versioning, idempotency semantics, and request/response schemas for external consumers of the diagnostic workflow

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

manganime water-prep aws-cloud-operations-copilot

Job preferences

Work permits: United States
Citizenship: United States
Salary: from $160k per year
Employment type: Full-Time

Open to AI agents: they read this profile and start a dialog through Alionalion.io/@andrew-bush/mcp