01
Extract
Structure the job posting
Turn an unstructured JD into company, role, technical skills, target contact type and a searchable outreach path.
OUTPUTStructured job metadata
I turned a repetitive job-search workflow into a human-in-the-loop automation system, combining structured data extraction, rules-based fit assessment, application prioritisation, tailored resumes, outreach and follow-up.
A job search contains a surprising amount of operational work: extracting requirements, comparing experience, deciding how much to tailor an application, finding the right contact, preparing outreach and remembering to follow up.
Instead of treating every vacancy as a separate task, I decomposed the workflow into repeatable stages and connected those stages through structured data and decision rules.
Turn an unstructured JD into company, role, technical skills, target contact type and a searchable outreach path.
Check dealbreakers first, then assess requirements against an authoritative candidate experience pool and produce a fit score.
The score feeds a defined application strategy: skip, apply cautiously, apply, or prioritise, with different tailoring depth.
Use the selected strategy to tailor the professional title, summary, competencies and relevant experience without inventing evidence.
Carry the structured job record into the tracker, then connect the vacancy with a relevant LinkedIn or email contact and tailored message.
Track contact status, follow-up dates, messages and application status so the workflow continues after the Apply button.
The fit assessment separates judgement from the downstream action. Requirements are assessed against the candidate profile, then the result is mapped to a predefined application strategy.
This prevents the workflow from becoming a black-box “apply or reject” machine. The system reduces repetitive analysis while keeping the final positioning and application decision human.
The score informs the next step, but does not become the final application decision. A defined strategy layer controls how much effort to invest.
AI handles extraction, classification and drafting. Judgement, positioning and relationship-building remain human responsibilities.
Company, role, skills, contact type and application metadata are standardised before downstream workflows use them.
High-fit roles receive deeper tailoring and targeted outreach. Weaker opportunities require less manual effort.
The workflow is designed around evidence from the candidate experience pool, so tailoring does not invent experience simply to match a JD.
The process continues after application submission through outreach, follow-up dates, response tracking and status updates.
I was spending too much time managing the mechanics of the search and not enough time on the parts that required judgement. Building the workflow became an opportunity to apply the same principles I use professionally: structure the process, make the data usable, identify bottlenecks, automate repetitive steps and create visibility over what needs attention next.