DeepCandidate

The description states that DeepCandidate is a local-first, open-source AI career memory and job application agent. It allows candidates to upload CVs (PDF/DOCX), paste project notes, add stories, or answer questions about their experience.

It uses:

• GPT-5.6 via OpenAI API for extracting structured evidence from inputs
• Text embedding (text-embedding-3-small) for semantic search
• A SQLite database with Drizzle ORM to store career data and application drafts
• Tools like Mammoth.js, unpdf, and vitest for document parsing and testing

Key features include:

• Structured extraction of facts including projects, challenges, decisions, results, skills, lessons learned, audits, etc.
• Semantic retrieval of relevant experience based on role context
• Approval flow where only approved evidence is used in generation
• Generation of ATS-readable CVs tailored to a specific role
• Preservation of source quotes and proof links
• Detection of conflicting information or missing context
• Refusal to generate fabricated experience

The system is designed to be local-first, meaning all data stays on the user’s device, and no external vector database is used.

The author positions DeepCandidate as:

“Same experience. Different story.”

This suggests that while the facts of a candidate's work remain unchanged, the way those experiences are framed can vary depending on the role they're applying for.

It is described not just as another CV generator but as an evolving digital memory of how one works, thinks, decides, fails, learns, and improves.

The author frames this as:

“I believe hiring in the AI era will become much deeper, and probably a little Black Mirror.”

This implies a vision of future hiring being more dynamic and personalized through AI-driven storytelling rather than static documents.

However, there is no evidence that DeepCandidate has moved beyond its initial concept or gained traction with users or partners. The positioning remains aspirational and self-reported.

The author describes the target customer as:

“A professional who wants to present their experience in a way that adapts to different roles.”

They also note:

“My experience can support applications for crypto investigation, Solidity development, product ownership, and now AI roles.”

This indicates an early-stage audience likely composed of:

• Early-career professionals
• Developers or technical roles requiring storytelling around past projects
• Job seekers in fast-moving fields like tech, crypto, or AI

However, the description does not provide any evidence of actual customer segments, personas, or usage patterns beyond the author's own use case.

There is no evidence provided about:

• Revenue streams
• Pricing models
• Monetization strategy
• Customer acquisition costs
• Any commercial relationships or partnerships

The project is described as an open-source tool built during a hackathon, with no indication of any business model beyond its creator’s personal experimentation.

The technical stack includes:

• Next.js, React, TypeScript
• Tailwind CSS for styling
• Drizzle ORM + libsql for local database storage
• OpenAI API (GPT-5.6 and text-embedding-3-small)
• Zod for schema validation
• Vitest for testing
• Mammoth.js, unpdf for document parsing

Key delivery signals:

• Local-first architecture with no cloud dependencies
• Strict schema validation of AI outputs
• Semantic search using local embeddings
• No vector database used in MVP
• Separation between AI generation and retrieval logic
• Versioned prompts
• Provider abstraction layer to allow future switching (e.g., Ollama)

The author emphasizes:

“I deliberately avoided LangChain, multi-agent frameworks, authentication, and a vector database to keep the product understandable and focused on one complete loop.”

This suggests a focus on simplicity and clarity over complexity.

There is no evidence of:

• Revenue
• Customers or users
• Adoption metrics
• Product usage data
• Any form of traction beyond the author’s own development

The project is explicitly described as a hackathon submission, indicating it has not yet entered any market or product maturity phase.

There are no references to competitors in the description. The author does not mention:

• Existing tools for CV generation
• Career memory or storytelling platforms
• AI-powered job application agents
• Other career-focused AI tools

Thus, there is no evidence of competitive positioning or awareness of existing solutions.

Several risks and red flags are implied by the description:

1. No commercial traction: The project is described as a hackathon submission with no evidence of real-world usage.
2. Single founder: Only one team member is mentioned, which may limit scalability or execution capability.
3. Unproven market demand: No indication that there is a market need for this type of tool beyond the author’s personal experience.
4. Limited product maturity: The system is described as an MVP with no production deployment or user feedback loops.
5. Self-reported claims only: All descriptions are self-reported and unverified; no third-party validation exists.

1. What specific problems do you see in current job application processes that your tool addresses?
2. Have you tested DeepCandidate with real users outside of yourself? If so, what were the results?
3. How do you plan to monetize or scale this product beyond a hackathon prototype?
4. What is the timeline for moving from MVP to a production-ready version?
5. Are there any legal or privacy considerations around storing personal career data locally?
6. Do you have plans to integrate with existing job platforms or HR systems?
7. How do you intend to ensure consistent quality of AI-generated content across different roles and contexts?