Strata
Treat job search as a strategic campaign, not a checkbox exercise.
What it is
Strata is an autonomous AI job-search operating system. Instead of scrolling job boards, you define your profile and target companies, then Strata's agents handle the rest: scanning career pages daily, detecting new listings, scoring them against your profile, and surfacing the matches worth your attention, with explanations of why each role matches and where the gaps are.
Every application requires your approval. Strata optimizes for signal quality, not spray-and-pray volume.
Design principles
| Principle | What it means |
|---|---|
| Precision over volume | Every application requires human approval. The system surfaces fewer, better opportunities, not more noise. |
| Agent autonomy within governance | Each agent has access to only what it needs. The Scraper Agent can't modify your profile. The Profile Agent can't submit applications. Enforced at the database level, not by convention. |
| Cost-conscious by default | A three-layer deduplication funnel eliminates 95%+ of duplicate listings before any expensive AI scoring happens. |
| Observable always | Every agent decision is logged. Ask "why did this job score 82?" and get the full reasoning chain. |
The eight agents
| Agent | What it does | Your involvement |
|---|---|---|
| Profile | Converts your resume into a rich Candidate Intelligence Profile (CIP), with full version history | You review and approve all identity data |
| Target | Manages your target company list with priority tiers (Tier 1 to Tier 3) | You define your strategy and company set |
| Scraper | Daily career-page sweeps across all your target companies | Runs automatically; auto-promotes validated results |
| Match | Two-stage AI scoring: vector similarity first, then deep LLM analysis for borderline cases | Runs automatically; surfaces results with confidence tiers |
| Apply | Generates tailored resumes and cover letters for high-match roles | You always review before anything is sent |
| Skills | Gap analysis against your highest-scoring matches: what skills would unlock more opportunities | Runs automatically from active match data |
| Network | Contact intelligence and outreach-timing recommendations | You verify imported contacts |
| Brand | LinkedIn content and portfolio positioning aligned to your target roles | You approve all published content |
How it works
What makes it different
You control every decision
Strata never auto-submits applications, never modifies your profile without approval, and never sends anything on your behalf. Governance is enforced at the infrastructure level.
The system explains itself
Every match score includes a reasoning trace: which requirements aligned, which gaps exist, and what's missing from your profile. No black boxes.
Signal over noise
Most job-search tools optimize for volume. Strata optimizes for relevance. The dedup pipeline means you see each opportunity once; scoring tiers separate strong matches from weak ones before they reach your queue.
Efficient by design
The three-layer deduplication funnel filters the vast majority of listings before any AI call. You can also run it fully local via Ollama.
Status
Alpha. The core discovery and matching pipeline is functional and in active use. The Next.js dashboard handles profile management, job review, target tracking, and application management. Notification channels and public release are in progress.
Open-source libraries
Two components of Strata are available as standalone Python packages, stateless and MIT-licensed. They return structured data; Strata decides what to persist and how to govern it.
| Package | Purpose | Install |
|---|---|---|
| strata-harvest | Career-page scraping across 50+ ATS providers | pip install strata-harvest |
| strata-match | Two-stage vector + LLM job-to-profile matching | pip install strata-match |
The Strata core application is proprietary. The strata-harvest and strata-match libraries are open-source and available on PyPI.