TalentAST

Inspiration

Every hiring cycle, thousands of qualified developers get rejected by automated Applicant Tracking Systems (ATS) without clear feedback. At the same time, generic AI resume tools attempt to solve this by generating superficial buzzwords that hiring managers discard immediately.

We wanted to treat technical career alignment not as a copy-editing task, but as a software compilation and code analysis problem: parsing unstructured job requirements into a structured Abstract Syntax Tree (AST), computing a deterministic semantic diff, and generating verifiable proof-of-work project roadmaps for genuine skill gaps.

Mathematical Formulation & Scoring

TalentAST models a technical role and candidate background as hierarchical competency trees:

$$\text{AST}_{\text{JD}} = { (c_i, w_i, s_i) \mid c_i \in \text{Competencies}, w_i \in, s_i \in \text{Seniority} }$$

$$\text{AST}_{\text{Candidate}} = { (c_j, e_j) \mid c_j \in \text{Skills}, e_j \in \text{Evidence} }$$

The Qualification Parity Index ($\mathcal{P}$) is deterministically computed over all required nodes:

$$\mathcal{P} = \left( \frac{\sum_{i} w_i \cdot \text{MatchScore}(c_i, \text{AST}{\text{Candidate}})}{\sum{i} w_i} \right) \times 100\%$$

Where the semantic delta represents the exact disqualification risk:

$$\Delta_{\text{Gap}} = \text{AST}{\text{JD}} \setminus \text{AST}{\text{Candidate}}$$

What It Does

1. AST Parsing: Deconstructs raw job descriptions and candidate resumes into structured syntax nodes (Hard Requirements, Tooling, Concurrency, Architecture, Seniority).
2. Visual Git-Style Diff: Evaluates candidate experience with three distinct semantic states:
   • + [MATCHED] (Green): Verified skills with existing production evidence.
   • ~ [ADJACENT] (Yellow): Adjacent competencies that can be truthfully realigned using quantified impact metrics.
   • - [CRITICAL GAP] (Red): Hard technical gaps that trigger automated screening filters.
3. ATS-Optimized Tailored Resume: Exports a clean, ATS-compliant Markdown/text resume that highlights relevant adjacent experience without fabricating skills.
4. 7-Day Proof-of-Work Bridge: For critical gaps, generates an actionable 7-day mini-project blueprint (architecture summary, deliverable checklist, and verified links to official documentation) so candidates can prove competence on GitHub.

How We Built It

• Frontend & UI: Built with Next.js (App Router), TypeScript, and Tailwind CSS, featuring an Analog Static Noir high-contrast visual interface.
• AST Compiler Pipeline: Multi-stage LLM orchestration utilizing strict Zod JSON schema enforcement to ensure structured, deterministic output.
• Deterministic Resource Anchoring: Integrated a verified registry of technical documentation and reference implementations (Kubernetes, Redis, Go, Docker) to eliminate broken or hallucinated links.

Challenges We Ran Into

• Eliminating Hallucinations in Learning Links: Ensuring that recommended technical tutorials and architectural references strictly pointed to valid official documentation rather than generated URLs.
• Truthful Alignment vs. Keyword Stuffing: Fine-tuning prompts and extraction schemas to restructure genuine adjacent experience without fabricating credentials.

What We Learned

Treating career documents as structured Abstract Syntax Trees provides far higher evaluation accuracy and actionable clarity than conversational chatbots. When technical gaps are clearly diagnosed, candidates can build real proof of work rather than relying on buzzwords.

What's Next for TalentAST

• Automated GitHub profile scanning to extract commit histories, repository architectures, and test coverage directly into the Candidate AST.
• One-click GitHub repository template generation for 7-day proof-of-work project blueprints.

Built With

• css
• javascript
• typescript