Hi I’m Wes Porter, an analytics engineer, geospatial dev, and past-time photographer currently based in Nashville, Tennessee (but you might find me at a coffeeshop in NYC, Austin, or Chicago depending on the time of year).

Helping Jobseekers at Indeed

Professionally I’ve been on Indeed’s Location Data team since 2018.

In my current role I

• build and maintain production data pipelines (think data engineering specific to location data)
• create internal apps to optimize the team’s workflows (Flask apps to make updating databases easier, interactive maps, etc)
• develope Python libraries
• design and launch dashboards (KPI monitoring and stakeholder comms)
• model and deploy a variety of location databases

I enjoy the type of work that requires creative thinking, a bit of resourcefulness, and a willingness to explore.

This usually involves finding a painstaking or complex task and turning it into a system the team can run, test, measure, and maintain.

One recent example: an AI pipeline that replaced a manual vendor review task. It cut about $90K a year in vendor spend and raised our throughput from roughly 120 to more than 4,200 jobs per day at about 95% accuracy.

For quick summaries of recent projects, see my portfolio.

How I approach my work

1. Find a slow or fragile workflow. A manual vendor task, an error-prone SQL process, an hours-long review, or an inherited pipeline nobody fully understands.
2. Learn enough of the system to solve the whole problem. That often means digging through code, databases, CI/CD, cloud infrastructure, scheduling, and monitoring, not just the analysis.
3. Build something practical. Usually a reusable, lightweight app, pipeline, library, dashboard, or validation step rather than a one-time answer.
4. Measure whether it worked. Cost, turnaround time, data quality, coverage, reliability, and analyst time.
5. Make the knowledge reusable. Process guides, validation checks, troubleshooting notes, and agent-readable project context, so people and AI agents can repeat the work safely.

Selected work

Production data and AI pipelines. I design and run scheduled pipelines that ingest data, apply rules or models, validate results, and publish reusable outputs. The AI location extraction pipeline combines GPT-4o-mini with rules and a TF-IDF classifier, and publishes results the team uses to track precision and recall.

Internal applications and self-service tools. I build tools that let analysts do complex work without hand-written SQL or one-off setup: a validated editor for our places database, a Streamlit app for location database updates and SQL validation, self-service regression testing, and dashboards such as the geocoding credit dashboard.

Location-data quality and safety. Database changes can shift how jobs geocode and how searches resolve. I wrote SQL validation checks that reject unsafe changes (invalid hierarchies, duplicates, bad coordinates, orphaned aliases) before they ship, and I taught an AI agent to review regression reports the way an analyst would as an additional stop-gap.

Location datasets. I led the technical work to build our transit db for the United States, Germany, and Japan: about 430,000 station records. A Python ETL and conflation workflow reduced thousands of possible manual comparisons to roughly ~250. In 2025 I built a transit updater and used it to refresh the Japan data, adding 27 stations, 2 lines, and 3 companies in a matter of seconds with zero service disruption using this method.

Measurement and decision support.

• Took one large applicant-tracking feed from roughly 20% to 90% of jobs with street addresses.
• Led coverage and data qualtiy analyses for our Geocoding vendors, making presentations to our leadership supported by custom dashboards, offering insights and recommendations.
• Evaluated two point-of-interest vendors with an AI-assisted workflow, cutting about a week of analysis to a 40-minute session
• Led a commute-time proof of concept with the open-source Valhalla routing engine that cut test runs from a full day to minutes.

Inherited and legacy systems. I often maintain systems after their original owners move on, and sometimes the right move is to retire them. I led the shutdown of a legacy Django application and its infrastructure, cutting cloud cost and maintenance while keeping the useful capabilities in smaller tools.

A favorite hackathon project

At an Indeed hackathon I presented GeoAmenities, a map-based job search prototype that scored each job by the transit stops and routes nearby, to the person who is now Indeed’s CEO.

Career arc

• 2018–2020: Geographic analysis, regression review, and geocoding quality, increasingly replacing manual data preparation with scripts.
• 2021: Measurement and automation: coverage dashboards, safe SQL generation, and helping colleagues with difficult data problems.
• 2022–2023: Data engineering: large transit datasets, a Python ETL pipeline, and moving applications and databases to AWS with Terraform.
• 2024–2025: Owning team systems: self-service regression workflows, dashboards, repository maintenance, and retiring legacy applications.
• 2025–2026: Production AI: LLM extraction, feedback classification, labeling workflows, and documenting how teams can make repositories clear to AI agents.

Elsewhere

Email · GitHub · LinkedIn