AI HR and Recruitment Automation

def screen_resume(resume_text, job_requirements, model="deepseek"): """Parse resume and score candidate fit against job requirements.""" prompt = f"""You are an experienced technical recruiter. Screen the following candidate against the job requirements. Job Requirements: {chr(10).join([f"- {r}" for r in job_requirements])} Candidate Resume: {resume_text[:8000]} Provide structured analysis: 1. CANDIDATE PROFILE: - Years of relevant experience - Highest relevant education - Current/most recent role and company - Key technical skills identified - Notable achievements 2. REQUIREMENTS MATCH: For each job requirement, assess: - Status: STRONG_MATCH | MATCH | PARTIAL_MATCH | NO_MATCH - Evidence from resume - Confidence: HIGH | MEDIUM | LOW 3. OVERALL SCORE: - Fit score (0-100) - Experience level: ENTRY | MID | SENIOR | LEAD | EXECUTIVE - Recommendation: STRONG_RECOMMEND | RECOMMEND | CONSIDER | PASS 4. RED FLAGS (if any): - Employment gaps - Job hopping - Missing critical qualifications - Inconsistencies 5. INTERVIEW_FOCUS_AREAS: - Top 3 areas to probe in interview - Suggested interview questions 6. DIVERSITY_NOTE: - Flag if candidate brings underrepresented perspective - Note: Do NOT let this influence the technical score Output as JSON. Be objective and evidence-based.""" return call_llm_api(prompt, temperature=0.2, max_tokens=2000, response_format="json")

import pandas as pd import asyncio async def batch_screen_resumes(resumes_df, job_requirements, top_n=20): """Screen multiple resumes and rank candidates.""" results = [] semaphore = asyncio.Semaphore(10) async def process_one(row): async with semaphore: result = await screen_resume_async( row['resume_text'], job_requirements ) return { "candidate_id": row['id'], "name": row['name'], "score": result.get('overall_score', 0), "recommendation": result.get('recommendation', 'PASS'), "details": result } tasks = [process_one(row) for _, row in resumes_df.iterrows()] results = await asyncio.gather(*tasks) # Sort by score descending results.sort(key=lambda x: x['score'], reverse=True) # Return top candidates top_candidates = [r for r in results[:top_n] if r['recommendation'] in ['STRONG_RECOMMEND', 'RECOMMEND']] return { "total_screened": len(results), "top_candidates": top_candidates, "strong_recommends": len([r for r in results if r['recommendation'] == 'STRONG_RECOMMEND']), "recommends": len([r for r in results if r['recommendation'] == 'RECOMMEND']), "considers": len([r for r in results if r['recommendation'] == 'CONSIDER']) }

def generate_job_description(role_title, team_info, requirements, tone="professional", model="glm"): """Generate optimized, inclusive job description.""" prompt = f"""Write a compelling job description for the following role. Role: {role_title} Team/Company: {team_info} Requirements: {chr(10).join([f"- {r}" for r in requirements])} Requirements for the job description: - Use inclusive language (avoid gendered terms, unnecessary degree requirements) - Focus on outcomes and impact, not just responsibilities - Include growth opportunities and learning path - Mention team culture briefly - Include compensation range placeholder [RANGE] - Include EEO statement - Length: 300-500 words - Tone: {tone} Structure: 1. Hook/Why this role matters (1-2 sentences) 2. What you'll do (5-7 bullet points) 3. What we're looking for (must-haves and nice-to-haves separated) 4. What you'll learn/grow into 5. About the team 6. Benefits/perks placeholder 7. EEO statement Avoid: laundry lists of every possible skill, vague buzzwords without specifics, requirements that could discourage diverse candidates.""" return call_llm_api(prompt, temperature=0.5, max_tokens=1200)

def generate_outreach_message(candidate_profile, role, company_info, outreach_type="linkedin"): """Generate personalized candidate outreach message.""" templates = { "linkedin": "Write a concise LinkedIn connection request + message", "email": "Write a professional recruitment email", "referral": "Write a message asking for a referral introduction" } prompt = f"""{templates.get(outreach_type, templates['email'])}. Candidate Profile: - Current role: {candidate_profile.get('current_role', 'Unknown')} - Company: {candidate_profile.get('company', 'Unknown')} - Background: {candidate_profile.get('background', 'Unknown')} - Notable achievements: {candidate_profile.get('achievements', 'Unknown')} Role: {role['title']} at {company_info['name']} Role highlights: {role.get('highlights', '')} Company: {company_info['description']} Requirements: - Personalize based on candidate's background - Mention 1-2 specific things from their profile - Explain why this role fits their trajectory - Keep it concise ({'300 chars' if outreach_type == 'linkedin' else '150 words'}) - Include clear call-to-action - Professional but warm tone - Do NOT use generic templates Output only the message text.""" return call_llm_api(prompt, temperature=0.6, max_tokens=400)

def generate_interview_questions(role, seniority, competencies, interview_type="technical"): """Generate tailored interview questions.""" prompt = f"""Generate a structured interview question set. Role: {role} Seniority: {seniority} Interview Type: {interview_type} Key Competencies: {', '.join(competencies)} For each competency, provide: 1. One behavioral question ("Tell me about a time when...") 2. One situational question ("How would you handle...") 3. For technical roles: One technical problem or case study For each question include: - What to listen for in the answer - Red flags in responses - Strong answer indicators Generate 8-12 questions total. Include scoring rubric (1-5 scale) for each competency. Format as structured JSON.""" return call_llm_api(prompt, temperature=0.4, max_tokens=2000, response_format="json") def score_interview_response(question, response, competency, rubric): """Score a candidate's interview response.""" prompt = f"""Score the following interview response against the rubric. Question: {question} Competency Being Assessed: {competency} Rubric: {rubric} Candidate Response: {response} Provide: 1. Score (1-5) 2. Justification for score 3. Key strengths demonstrated 4. Areas of concern (if any) 5. Follow-up questions to probe deeper 6. Overall impression: EXCEEDS | MEETS | BELOW_EXPECTATIONS Be objective and evidence-based. Consider cultural and language differences.""" return call_llm_api(prompt, temperature=0.2, max_tokens=800)

def analyze_employee_feedback(feedback_texts, department="general"): """Analyze employee feedback for sentiment and themes.""" all_feedback = "\n\n---\n\n".join([f"Feedback {i+1}: {text}" for i, text in enumerate(feedback_texts)]) prompt = f"""Analyze the following employee feedback for {department} department. Feedback: {all_feedback} Provide: 1. OVERALL_SENTIMENT: Positive/Neutral/Negative with score (-1.0 to +1.0) 2. KEY THEMES: Top 5 themes mentioned (positive and negative) 3. SENTIMENT_BY_TOPIC: - Management/Leadership - Compensation/Benefits - Work-Life Balance - Career Growth - Team Culture - Tools/Resources 4. URGENT_ISSUES: Any issues requiring immediate management attention 5. TRENDS: Patterns across multiple feedback items 6. ACTIONABLE_RECOMMENDATIONS: Specific, prioritized actions for leadership 7. ANONYMIZED_QUOTES: 3-5 representative quotes (anonymized) that capture key themes Maintain strict confidentiality. Do not attempt to identify individuals from writing style. Output as JSON.""" return call_llm_api(prompt, temperature=0.2, max_tokens=2000, response_format="json")

def generate_onboarding_plan(employee_profile, role, team, start_date): """Generate personalized 30-60-90 day onboarding plan.""" prompt = f"""Create a comprehensive onboarding plan. New Hire: {employee_profile['name']} Role: {role['title']} Team: {team['name']} Start Date: {start_date} Background: {employee_profile.get('background', 'Not specified')} Experience Level: {employee_profile.get('level', 'Mid-level')} Generate: 1. PRE-START (Before Day 1): - Welcome email content - Access/setup checklist - First-day schedule 2. WEEK 1 (Days 1-5): - Daily agenda - Key meetings to schedule - Initial learning objectives - Buddy/mentor assignment 3. 30-DAY PLAN: - Key objectives (3-5 specific, measurable goals) - Training modules - Early wins to aim for 4. 60-DAY PLAN: - Deeper dive projects - Cross-functional introductions - Feedback checkpoints 5. 90-DAY PLAN: - Independent project ownership - Performance review prep - Career path discussion 6. RESOURCES: - Key documents to read - Tools to master - People to meet Make it specific and actionable. Personalize based on experience level.""" return call_llm_api(prompt, temperature=0.4, max_tokens=2500)