Sakto ka

Find jobs that truly match your skills — without the guesswork.

Comment

Inspiration

Job hunting today feels like shouting into an algorithmic black hole. Job seekers—especially fresh graduates and career switchers—spend hundreds of hours applying to roles blindly without knowing whether they are genuinely qualified or why they get ghosted.

We built Sakto ka ("You're a match!" / "Spot on!") to eliminate the guesswork. We wanted to take the chaos out of job applications and turn the entire search into a transparent, feedback-driven pipeline where candidates know exactly where they stand before they even hit "Apply."

What it does

Sakto ka is an intelligent job search and application pipeline platform that bridges the gap between candidate qualifications and job postings.

🎯 Smart Match & Qualification Scoring Engine: Evaluates job descriptions against candidate profiles across 6 weighted dimensions (Technical Skills, Role Compatibility, Experience, Education, Work Setup, and Compensation) to output clear APPLY, REVIEW, or SKIP badges with breakdown scores. 🧹 Instant Job URL Ingestion: Cleans tracking parameters and strips noisy boilerplate from URLs using automated web extractors (selectolax & BeautifulSoup). 📋 Interactive Application Kanban: Visual drag-and-drop pipeline organizing applications from Saved → Applied → Screening → Technical Round → Final Round → Offer. 🎙️ AI Interview Prep Studio: Generates role-specific technical questions, STAR-method behavioral response blueprints, and intelligent reverse-interview questions to ask employers. 📬 Zero-Ghosting Follow-Up Assistant: Auto-schedules 5-day and 10-day status checkpoints with one-click customizable follow-up email drafts. 📊 Skill Gap & Funnel Analytics: Pinpoints recurring missing skills across target jobs so candidates know exactly what to learn next.

How we built it

We designed Sakto ka with a modern, high-performance full-stack architecture:

Backend: Built with FastAPI (Python 3.11+) and SQLAlchemy for lightning-fast asynchronous REST APIs. Frontend: Built with Next.js 15 (App Router), React 19, and TypeScript styled with a sleek, responsive design system. Multi-Provider AI Architecture: A flexible AI abstraction factory supporting Google Gemini, Groq, NVIDIA NIM, OpenAI, and Zhipu AI, backed by a 100% offline, zero-dependency Deterministic Local Heuristic Engine for instant fallbacks. Data Processing & Parsing: High-speed HTML text sanitization and metadata extraction powered by selectolax and BeautifulSoup4. Storage: Async SQLite (aiosqlite) for lightweight local persistence, architected to scale seamlessly to PostgreSQL. Challenges we ran into Handling Messy Job Postings: Web pages from various job boards have inconsistent DOM structures, dynamic JS rendering, and trackers. We solved this by developing a multi-tiered extraction pipeline with heuristic sanitization and fallback LLM parsing. Deterministic & Fair Scoring: Pure LLM scoring can hallucinate or vary between runs. We paired the AI with a strict mathematical weighting model (e.g., Skills 35%, Experience 15%) so scores remain explainable, reproducible, and actionable. Multi-Model Provider Resilience: Ensuring zero downtime when API keys or provider limits fluctuate required building an extensible provider factory with graceful fallbacks. Accomplishments that we're proud of ✨ Zero-Key Offline Mode: Building a deterministic matching and parsing engine that works out-of-the-box even without paid API keys. ⚡ Sub-Second Job Analysis: Achieving blazing fast URL ingestion and match calculation. 🎯 Actionable Transparency: Replacing vague rejection anxiety with clear, actionable feedback—showing users exactly which skills to highlight and which gaps to bridge. 🛠️ End-to-End Workflow: Unifying job discovery, qualification matching, Kanban tracking, and interview preparation into a single cohesive interface.

What we learned

How to combine deterministic heuristics with generative AI to make evaluations trustworthy and reproducible rather than arbitrary. Structuring LLM outputs with strict JSON schemas and Pydantic validation to guarantee reliable UI rendering. Designing user-centric pipelines that reduce cognitive overload and burnout during stressful career transitions.

What's next for Sakto ka

🧩 Chrome Extension: One-click job clipping and on-page match analysis directly on LinkedIn, Indeed, and company career pages. 📄 Dynamic Resume & Cover Letter Tailoring: Generating ATS-optimized resume bullet points tailored to individual job requirements. 🤖 Mock Voice Interviewer: Interactive, real-time voice practice with instant AI feedback on communication and technical depth. 👥 Peer & Mentor Review Circles: Allowing mentors or peers to review application stages and share interview tips directly in the pipeline.

Built With

Share this project:

Updates

Submission history