ChatWith.work
AI-powered CareerOps & Job Search System
An AI-powered career operating system for resume optimization, career planning, job discovery, application automation, and application tracking.
01 / CONTEXT
The problem
Job applications are highly repetitive but not uniform. Different platforms expose different application schemas, dynamic fields, authentication states, and verification requirements. Automating the process therefore requires adaptable browser automation rather than a single rigid workflow.
02 / APPROACH
How I built it
Job listings are normalized and matched against user profiles. Application automation is delegated to platform-specific adapters capable of mapping structured application fields, adapting to selector changes, and pausing for human intervention when verification or authentication requirements cannot safely be automated.
03 / CONSTRAINTS
What had to hold
25+ job platforms Different application schemas Dynamic fields Authentication states Human verification 2FA Need for human oversight
04 / OBJECTIVES
Definition of done
• Automate resume tailoring • Improve ATS alignment • Provide career-path guidance • Identify skill gaps • Aggregate relevant jobs • Automate repetitive application steps • Track application progress
05 / THE SYSTEM
Architecture
Career intelligence layer Resume, career path, skills, and target-role information are represented as structured career data. Semantic matching Job descriptions are compared against candidate experience and career goals. Application adapters Individual job platforms are handled through modular automation adapters. Distributed workers Job discovery and applications execute asynchronously through background workers. Human-in-the-loop controls Verification, authentication, or exceptional cases can be handed back to the user.
06 / DECISIONS
Key decisions
Use modular platform adapters Why: Job sites differ significantly in application workflows and DOM structures. Use structured LLM form mapping Why: AI can help map normalized candidate information to varying application schemas. Retain human-in-the-loop fallback Why: Authentication and human verification cannot reliably be treated as ordinary automation steps. Separate job ingestion from application execution Why: Discovery and application workflows have different processing characteristics and failure modes.
07 / SHIPPED
Deliverables
• AI CV builder (in-progress) — Core capability of the product (see features). • Resume optimization (in-progress) — Core capability of the product (see features). • ATS optimization scoring (in-progress) — Core capability of the product (see features). • Career path planner (in-progress) — Core capability of the product (see features). • Skill gap analysis (in-progress) — Core capability of the product (see features).
08 / LESSONS
What I'd keep
• Automation alone does not produce better job-search outcomes. • Role-specific resume tailoring and customized positioning are essential. • Browser automation needs platform-specific adapters and human fallback paths. • High-quality semantic matching is more important than simply increasing application volume.
What changed
45 minutes → 1 click
Application workflow time
to date
// START A PROJECT
Building something with similar constraints?
If it's in your critical path and you'd rather not learn its failure modes live, let's talk about it.