Andrés Chávez

AI & Automation Engineer, Chilliwack, BC, Canada

I build LLM systems that run in production — with human approval gates, fact-checking, and the boring operational plumbing that keeps them honest.

Email LinkedIn GitHub Download PDF

andres@streamlite.ca · https://www.linkedin.com/in/jachaveza/ · https://github.com/jachaveza26 · https://github.com/jachaveza26/portfolio

Portrait of Andrés Chávez

About

AI & Automation Engineer and founder of Streamlite Technologies, building LLM systems that run in production for clients in Canada, the US and Mexico: integrations with APIs that fight back, human approval gates so nothing publishes unreviewed, and the operational plumbing that keeps both honest.

Six years of data analytics underneath — Power BI, DAX, SQL — which is why I care what a number means before I automate it. Before that, a decade of systems work in Mexico: software engineering, SAP ABAP, web applications and end-user support.

I own the whole line: discovery with the client, architecture, build, deploy, and the 2 a.m. alert. Bilingual in Spanish and English. Open to AI / Automation Engineer and Forward Deployed Engineer roles.

Experience

Founder & AI/Automation Engineer

Streamlite Technologies, Chilliwack, BC

– Present

AI automation consultancy for small and mid-size businesses across Canada, the US and Mexico. Sole engineer on every engagement: discovery, architecture, build, deploy, billing and on-call.

  • Built and run GBP Manager, a multi-tenant LLM SaaS for Google Business Profiles; nothing publishes without a per-item human approval record. Google API production access approved after two rejections.
  • Automated Canadian immigration (IRCC) PDF forms from one canonical schema using Claude tool use and XFA XML; in production for an immigration consultancy.
  • Delivered ScopeForge, a bid-intelligence platform for a millwork subcontractor: AI scope extraction from bid packages, BuildingConnected sync, DOCX proposals and an automated 3-step follow-up.
  • Own a ~30M-row Power BI model and a PLD/AML transactional monitoring layer (fraud detectors, per-merchant thresholds, audited retention) for a Mexican POS distributor.
  • Shipped Relay, social and local-SEO automation on the Meta Graph API at Standard Access, deployed per client on isolated branches; one paying client.
  • Built a Microsoft licence drift detector as a paid-evaluation prototype for an MSP: 100% recall, zero false positives on a ground-truth set.

Case studies: GBP Manager · IRCC Form Automation · ScopeForge · Prosepago · Relay · Licence Reconciliation

Computer System Analyst

Fernando Torres Immigration, Vancouver, BC

–

  • Implemented and maintained the CRM system (Pipedrive), ensuring data accuracy and consistency.
  • Led website functionality updates and content management for immigration programs and services.
  • Developed landing pages to optimize lead generation and conversion.
  • Analyzed data to generate Power BI reports and insights for decision-making on immigration client needs.

Data Analyst

Agile Thought Digital Solutions, Mexico

–

  • Designed and developed Power BI dashboards on sales performance and customer behaviour for the retail industry.
  • Implemented advanced data visualizations in Power BI for comprehensive analysis.
  • Ran training sessions on effective use of Power BI for the data analytics team.

Data Analyst

Conduent de México, Mexico

–

  • Wrote and optimized SQL statements for more efficient data analysis.
  • Designed customized data collection models for visualization tasks.
  • Created data analysis tools for easy querying and insight generation.
  • Implemented advanced data visualizations in Power BI dashboards.

Business Data Analyst (part-time)

Prosepago, Mexico

–

Part-time engagement with the POS distributor that re-engaged Streamlite in 2025 for its BI and PLD/AML stack (see above).

  • Implemented ActiveCampaign as the CRM and trained employees on it for improved data management.
  • Implemented advanced data visualizations in Power BI dashboards.

Earlier career

Desarrollos y Servicios Computacionales · LANIX, Mexico

–

Computer systems analyst at DSC (2013–2020): end-user support across operating systems, software and hardware solutions for restaurant-industry clients, website build and monitoring. Software engineer at LANIX (2009–2013): server and virtual-machine management, web applications and database design, SAP ABAP (Z tables, Smart Forms, user exits, ALV reports).

Selected projects

One case study per project, written for another engineer, in the portfolio repository.

GBP Manager

Multi-tenant AI SaaS that audits Google Business Profiles, drafts review replies and posts — every publish behind a human approval record.

  • Next.js
  • Python
  • Supabase
  • Stripe
  • Railway
Production, pre-traction

IRCC Form Automation

Fills Canadian immigration PDFs from a canonical intake schema using LLM tool use and XFA XML manipulation.

  • FastAPI
  • Claude Tool Use
  • XFA XML
Production

ScopeForge / MCI Bids

Bid-intelligence platform for commercial construction subcontractors: AI scope extraction, proposal generation, pipeline tracking.

  • Python
  • Railway
  • BuildingConnected API
Production, 1 client

Relay

Multi-tenant social and local-SEO automation, deployed per client on isolated branches.

  • TypeScript
  • Supabase
  • Meta Graph API
  • DataForSEO
Production, paying client

Prosepago

~30M-row Power BI model for a Mexican POS distributor, plus PLD/AML transactional monitoring and open-data prospecting.

  • Power BI
  • DAX
  • Python
  • PostgreSQL
Production

Also built

Eden Estimator

The repo where my engineering practices live: Vitest suite, Docker, Caddy, pnpm monorepo, sliced delivery.

  • TypeScript
  • Vitest
  • Docker
  • Supabase
In development

Licence Reconciliation

Microsoft licence drift detector built as a paid-evaluation prototype — 100% recall, zero false positives on a ground-truth set.

  • Python
  • FastAPI
  • Microsoft Graph
Prototype, delivered

Voice Agents

Retell and VAPI voice agents embedded in a Next.js product surface.

  • Next.js
  • Retell AI
  • VAPI
Production

CFSF Loan Automation

Nine-deliverable automation architecture for a non-profit lender, designed under a hard no-PII constraint.

  • FastAPI
  • AI agents
  • DocuSign (Phase 2)
Designed, not built

Content Pipeline

Near-fully automated Spanish video pipeline with data-driven niche validation and a tiered model-cost strategy.

  • Python
  • SQLite
  • Streamlit
  • multi-model video APIs
Production

Anota

Native Android note app — my first mobile project, built to find out where AI assistance stops working.

  • Kotlin
  • Gradle
In development

Skills

Technical

Languages
Python · TypeScript · SQL / PLpgSQL · Kotlin · DAX · R
AI / LLM
Claude API & tool use · Gemini · OpenAI · MCP servers · OpenRouter · multi-agent orchestration · guardrail design · fact-checking · prompt versioning · Retell · VAPI · ElevenLabs
Backend
FastAPI · Supabase (Postgres, auth, pg_cron) · Next.js API routes · n8n
Frontend
Next.js · React · Vite · Tailwind
Data / BI
Power BI · DAX Studio · Tabular ALM Toolkit · Power Query · Apify · Excel
APIs & integrations
Google Business Profile API · Microsoft Graph · Meta Graph (Standard Access) · WhatsApp Cloud API · BuildingConnected · Jobber GraphQL · Zoho · Stripe · Twilio · DataForSEO
Infra
Docker · Railway · Cloudflare Pages · GitHub Actions
Quality
conventional commits · sliced delivery
Security
OAuth 2.0 · per-tenant ownership validation

How I work

Client discovery and scoping
Runs paid discovery workshops credited against the build, so scope is defined before code is written.
Pricing and proposals
Prices and structures engagements — implementation fee plus retainer — writes the proposals and negotiates bounded concessions.
End-to-end ownership
Discovery, architecture, build, deploy, billing and the on-call pager, usually as the only engineer on the project.
Honest status reporting
Every case study ends with a status line, flattering or not; roadmap is labelled as roadmap in every deck.
Communicating to different audiences
Presents the same analysis twice a week: dense detail to the commercial team on Monday, one message per slide to executives on Thursday.
Bilingual delivery
Spanish with teams in Mexico, English with clients in Canada and the US; half the projects were scoped in one language and delivered in the other.
Stakeholder engagement
Maps automation proposals to what each stakeholder actually does, not to their title.
Attention to detail
Root-caused a company-wide BI outage to one duplicated name in a lookup table.
Training and enablement
Ran Power BI training for an analytics team and CRM onboarding for staff at two companies.

Education

  • Master's Degree in Big Data and Omnichannel — Esden Business School, Spain, 2020
  • Bachelor's Degree in Computer Science — Instituto Tecnológico de Hermosillo, Mexico, 2010

Certifications

  • Business Analyst Certificate — Camosun College, Canada, Mar 2025
  • Google Data Analytics Certificate — Esden Business School, Spain, 2020

Languages

Spanish (Native) · English (Fluent)