Better interviews. Deeper insights.
Smarter hiring decisions.
Obsium built an AI-powered Interview Copilot for ellow that joins candidate interviews, captures the conversation, and delivers structured insights, so interviewers stay focused on people and humans still make the final call.
The Client: An AI-Native Talent Agency
ellow is an AI-native talent agency that blends technology, AI, and human expertise to solve hiring challenges more effectively than traditional recruitment agencies.
Where conventional firms lean on one or the other, ellow brings together AI-driven sourcing, intelligent candidate evaluation, and human judgment. The result is faster, more accurate, better-informed hiring decisions.
To keep reinventing recruitment, ellow asked Obsium to build an AI-powered Interview Copilot: a second, AI-driven perspective that sits alongside human evaluation in every interview.
The Challenge: Great Interviews Ask a Lot of One Person
Hiring the right person takes more than checking technical skills. Interviewers also have to judge communication, problem-solving, relevant experience, and overall fit for the role. Traditional interviews make that harder than it should be.
Divided attention
Engaging the candidate, asking the right questions, taking notes, and evaluating answers, all at the same time.
Subjective evaluations
Two interviewers can hear the same answer and walk away with different conclusions.
Limited documentation
Important responses and observations slip through or never make it into the notes.
Hard-to-compare candidates
Inconsistent notes make objective side-by-side comparisons difficult.
AI-only interviews fall short
Fully automated interviews miss the context, adaptability, empathy, and professional judgment experienced interviewers bring.
The goal isn't to replace interviewers with AI. It's to combine human intelligence with AI-powered analysis to make better hiring decisions.
The Solution: Meet Interview Copilot
Obsium built Interview Copilot, an automated meeting assistant that joins candidate interviews, captures the conversation, and turns it into a structured interview intelligence report.
Interviewers stay fully focused on the candidate. AI takes care of recording, transcription, and post-interview analysis.
After every interview, the system analyzes the conversation and delivers a structured report: an additional AI perspective on the candidate's responses and demonstrated competencies.
Humans bring
- Experience
- Contextual understanding
- Professional judgment
AI brings
- Structured analysis
- Consistency
- Evidence-based insights
How Interview Copilot Works: From Calendar Invite to Hiring Decision
- 1
Interview scheduled
The interview is booked on Google Meet, and Interview Copilot picks up the meeting details.
- 2
Copilot joins the call
A browser-based bot joins the session, waits for admission if needed, and monitors the meeting from start to finish.
- 3
Conversation recorded
Once recording is authorized, audio is captured automatically. No extra tools for interviewers to manage.
- 4
AI transcription
When the interview ends, the recording becomes a searchable transcript.
- 5
Intelligent analysis
AI reviews the transcript and pulls out structured, job-relevant insights.
- 6
Report delivered
A structured report lands in the designated recipients' inboxes automatically.
- 7
Humans decide
Recruiters and hiring managers weigh the AI insights against their own observations and make the call.
What the analysis can cover
- Technical knowledge demonstrated
- Communication clarity and effectiveness
- Problem-solving approach
- Relevance of experience to the role
- Strengths and areas to probe further
- Evidence behind each assessment
- Suggested follow-up questions
Built to evaluate job-relevant evidence, not to make unsupported judgments about candidates.
AI provides the additional perspective. Humans make the final decision.
Key Technical Capabilities: Automated End to End, Built to Scale
Every stage runs on an automated, scalable architecture, designed to process interviews reliably at scale with minimal manual intervention.
Google Meet automation
- Playwright-driven Chromium browser automation
- Joins meetings automatically
- Handles waiting rooms and admission
- Monitors the full meeting lifecycle
Smart meeting monitoring
- Detects when a meeting ends
- Handles participant removal
- Detects empty rooms
- Enforces maximum meeting duration
Audio capture and processing
- PulseAudio-based audio capture
- FFmpeg recording
- Recording integrity validation
Workflow orchestration
- Celery-based background processing
- Automated transcription and analysis
- Retries for failed processing stages
- Resource cleanup and failure recovery
Automated report delivery
- Structured AI-generated reports
- Email distribution
- Processing-status tracking
The Technical Challenge: Automating a Live Interview
An interview isn't one task. Joining a scheduled Google Meet session, capturing audio, transcribing, analyzing, and generating a report all have to work together, every time.
Obsium built the meeting assistant on Playwright and Chromium to handle participation, with PulseAudio and FFmpeg managing audio recording. Celery and Redis run processing asynchronously, so transcription, AI analysis, and report delivery happen in the background. Lifecycle monitoring, retry mechanisms, and resource cleanup handle interruptions and failures on their own, cutting the need for manual intervention.
Cloud, DevOps and Scale: Cloud-Native on Amazon EKS
Obsium designed and deployed cloud-native infrastructure on Amazon EKS (Elastic Kubernetes Service) to run the AI applications and their background processing workloads.
Every application runs as a containerized workload managed by Kubernetes, giving the platform a scalable foundation for rising processing demand and future growth.
GitHub Actions CI/CD pipelines automate deployments, cutting manual effort and keeping every release consistent.
Together, Kubernetes orchestration and automated deployments give the platform an infrastructure built for scalability, reliability, and easier maintenance.
Business Value: What It Means for the Business
Better hiring decisions
Human expertise plus AI-generated insights for more complete candidate evaluations.
Better interviews
Interviewers focus on meaningful conversation, not recording and note-taking.
Consistent evaluation
Structured records make it easier to compare candidates on common, job-related criteria.
Less admin
Recording, transcription, analysis, and report delivery happen automatically.
Scalable recruiting
Review more interviews without a matching rise in admin work.
Decision confidence
An evidence-based second perspective that complements professional judgment.
Responsible AI and Human Oversight
Decision support, not a decision-maker.
- Interview Copilot supports hiring decisions. It never makes them.
- Human reviewers stay responsible for evaluating candidates and making the final call.
- Participant notice, recording consent, secure data handling, role-based access, and retention controls should be part of the solution.
- AI assessments should stay transparent, grounded in interview evidence, and open to human review.
The Outcome: Interviews Turned Into Actionable Hiring Intelligence
Interview Copilot is live. Obsium built and deployed it as a fully operational AI-powered application, and it now actively supports ellow's recruitment and HR team.
Interview participation, audio recording, transcription, AI analysis, and report delivery are all automated, taking the manual work out of interview documentation and post-interview processing.
Running on AWS EKS with automated GitHub Actions CI/CD pipelines, the application has a scalable foundation for growing recruitment workloads.
Structured, AI-generated insights that complement human judgment throughout the candidate evaluation process.
What the solution supports
- Better-informed hiring decisions
- More consistent candidate evaluations
- Improved interviewer productivity
- Faster post-interview reviews
- Scalable interview intelligence
- A stronger mix of human judgment and AI insight
What This Makes Possible
With Interview Copilot in place, ellow's interviewers can focus on the candidate instead of the notepad, and hiring managers get a consistent, evidence-backed view of every interview to weigh alongside their own judgment. More interviews can be handled without more admin.
The platform also sets a strong base for responsible growth. Consent and notice flows, role-based access, retention controls, and deeper analysis can all be layered on as ellow scales, without disrupting the live workflow.
Ready to put AI to work alongside your team?
Obsium builds AI-powered applications that take on the busywork, run reliably on Kubernetes, and keep people in charge of the decisions that matter.
Get the case study
Enter your details to download the PDF.
