Resume Screening Assistant

1import { relay } from "@relayplane/workflows";
2
3const result = await relay
4  .workflow("hiring-screen")
5
6  // Step 1: Parse resume into structured data
7  .step("parse-resume")
8  .with("openai:gpt-4o")
9  .prompt(`Extract structured data from this resume:
10
11{{resumeText}}
12
13Return JSON:
14{
15  "name": "...",
16  "contact": { "email": "...", "phone": "...", "location": "..." },
17  "summary": "2-3 sentence professional summary",
18  "experience": [
19    {
20      "title": "...",
21      "company": "...",
22      "duration": "months as number",
23      "responsibilities": ["..."],
24      "achievements": ["..."]
25    }
26  ],
27  "skills": {
28    "technical": ["..."],
29    "soft": ["..."],
30    "certifications": ["..."]
31  },
32  "education": [
33    {
34      "degree": "...",
35      "field": "...",
36      "institution": "...",
37      "year": "..."
38    }
39  ],
40  "totalYearsExperience": number
41}
42
43Extract exact information, don't infer.`)
44
45  // Step 2: Match against job requirements
46  .step("match-requirements")
47  .with("anthropic:claude-3.5-sonnet")
48  .depends("parse-resume")
49  .prompt(`Match candidate against job requirements:
50
51Parsed Resume: {{parse-resume.output}}
52
53Job Requirements:
54{{jobRequirements}}
55
56For each requirement, assess:
57- Meets: Yes/Partial/No
58- Evidence: specific resume content supporting assessment
59- Gap: what's missing (if any)
60
61Categories:
62- Required Technical Skills
63- Years of Experience
64- Education
65- Preferred/Nice-to-have Skills
66- Industry Experience
67
68Be objective and evidence-based.`)
69
70  // Step 3: Score candidate
71  .step("score-candidate")
72  .with("openai:gpt-4o")
73  .depends("parse-resume", "match-requirements")
74  .prompt(`Calculate match score:
75
76Resume: {{parse-resume.output}}
77Requirements Match: {{match-requirements.output}}
78
79Scoring weights:
80- Required skills: 40%
81- Experience level: 25%
82- Industry relevance: 15%
83- Education: 10%
84- Nice-to-haves: 10%
85
86Return:
87- Overall score (0-100)
88- Breakdown by category
89- Confidence level (high/medium/low)
90
91Thresholds:
92- 80+: Strong match
93- 60-79: Good match
94- 40-59: Partial match
95- Below 40: Not a match`)
96
97  // Step 4: Generate screening summary
98  .step("screening-summary")
99  .with("anthropic:claude-3.5-sonnet")
100  .depends("parse-resume", "match-requirements", "score-candidate")
101  .prompt(`Generate screening summary for recruiter:
102
103Candidate: {{parse-resume.output}}
104Requirements: {{match-requirements.output}}
105Score: {{score-candidate.output}}
106
107Format:
108## Candidate Overview
109- Name and current role
110- Total experience
111- Match score and recommendation
112
113## Strengths
114- Top 3-5 reasons this candidate fits
115- Specific achievements relevant to role
116
117## Concerns/Gaps
118- Missing requirements
119- Areas to probe in interview
120
121## Interview Questions
122- 3-5 targeted questions based on gaps or verification needs
123
124## Recommendation
125- ADVANCE / HOLD / PASS
126- Brief justification
127
128Objective, professional tone.
129Focus on job-relevant factors only.`)
130
131  .run({
132    resumeText: candidateResume,
133    jobRequirements: {
134      title: "Senior Software Engineer",
135      required: [
136        "5+ years software development experience",
137        "Proficient in TypeScript and React",
138        "Experience with PostgreSQL or similar RDBMS",
139        "Background in CI/CD and DevOps practices",
140      ],
141      preferred: [
142        "Experience with AI/ML integration",
143        "Open source contributions",
144        "Previous startup experience",
145      ],
146      education: "Bachelor's in CS or equivalent experience",
147    },
148  });
149
150// Save to ATS
151await updateATS({
152  candidateId: candidate.id,
153  screeningScore: JSON.parse(result.steps["score-candidate"].output).score,
154  screeningSummary: result.steps["screening-summary"].output,
155  status: result.steps["screening-summary"].output.includes("ADVANCE")
156    ? "Phone Screen"
157    : "Reviewed",
158});

1// Process application queue
2import { relay } from "@relayplane/workflows";
3
4async function processApplicationQueue(jobId: string) {
5  const applications = await getNewApplications(jobId);
6  const jobRequirements = await getJobRequirements(jobId);
7
8  const results = [];
9
10  for (const app of applications) {
11    const screening = await relay
12      .workflow("hiring-screen")
13      .run({
14        resumeText: app.resume,
15        jobRequirements,
16      });
17
18    const score = JSON.parse(screening.steps["score-candidate"].output);
19
20    results.push({
21      candidateId: app.candidateId,
22      score: score.score,
23      recommendation: score.recommendation,
24      summary: screening.steps["screening-summary"].output,
25    });
26  }
27
28  // Sort by score and return top candidates
29  const ranked = results.sort((a, b) => b.score - a.score);
30
31  // Auto-advance top scorers
32  const autoAdvance = ranked.filter(r => r.score >= 80);
33  for (const candidate of autoAdvance) {
34    await schedulePhoneScreen(candidate.candidateId);
35  }
36
37  return ranked;
38}

1// Greenhouse/Lever webhook handler
2app.post("/webhooks/greenhouse", async (req, res) => {
3  const { action, payload } = req.body;
4
5  if (action === "candidate.applied") {
6    const { candidate, application, job } = payload;
7
8    // Queue for screening
9    await queue.add("screen-resume", {
10      candidateId: candidate.id,
11      resumeUrl: application.resume.url,
12      jobId: job.id,
13    });
14  }
15
16  res.sendStatus(200);
17});
18
19// Queue processor
20queue.process("screen-resume", async (job) => {
21  const resume = await downloadResume(job.data.resumeUrl);
22  const requirements = await getJobRequirements(job.data.jobId);
23
24  const result = await relay
25    .workflow("hiring-screen")
26    .run({
27      resumeText: resume,
28      jobRequirements: requirements,
29    });
30
31  // Update Greenhouse
32  await greenhouse.candidates.addNote(job.data.candidateId, {
33    body: result.steps["screening-summary"].output,
34    user_id: BOT_USER_ID,
35  });
36});

1## Candidate Overview
2**Sarah Chen** - Staff Software Engineer at DataCorp
3**Experience:** 8 years in software development
4**Match Score:** 87/100 (Strong Match)
5**Recommendation:** ADVANCE to phone screen
6
7## Strengths
81. **Strong TypeScript/React experience** - 5 years building production React apps at scale (DataCorp processes 1M+ daily transactions)
92. **Relevant PostgreSQL expertise** - Led database optimization project reducing query times by 60%
103. **DevOps/CI-CD** - Implemented GitHub Actions pipeline, Kubernetes deployments
114. **Architecture experience** - Designed microservices architecture serving 10M users
125. **Startup background** - Previous role at early-stage startup (20 employees)
13
14## Concerns/Gaps
151. **AI/ML experience not mentioned** - Resume doesn't show direct AI work (but this is preferred, not required)
162. **Open source** - No public contributions visible
173. **Career gap** - 6-month gap in 2021 (may have explanation)
18
19## Interview Questions
201. "Tell me about your database optimization project at DataCorp. What specific techniques did you use?"
212. "Have you worked with any AI/ML technologies? Any interest in that area?"
223. "I see a gap in early 2021 - what were you doing during that time?"
234. "Describe a time you had to make a difficult architectural decision with trade-offs."
245. "What interests you about moving from a large company to our startup?"
25
26## Recommendation
27**ADVANCE** - Strong technical match with 8 years experience and relevant stack expertise. Verify culture fit and interest in AI during phone screen. Clarify career gap.