Actually no, the problem is that technology has removed all friction in the application process which means that anyone anywhere with any skills can apply to any jobs at no cost and no time investment. The result is talent acquisition teams flooded with thousands of applicants. They have no choice but to reply on dumb ATS filters which arbitrarily block applicants. We are providing an alternative, a way for candidates to showcase their skills through a video interview with an AI that understands them.
Thank you for the feedback, the problem at the moment is that it's not AI vs Humans, it's job seekers applying to hundreds of jobs and never hearing back from the recruiters. It's not AI vs Humans it's AI vs being ghosted, AI vs application void, AI vs keyword filters. Our platform allows people to showcase their skills at least rather than getting blocked by an ATS.
Hey HN!
We built an AI recruiting agent that conducts actual voice interviews with candidates over the phone. It sources from 220M+ profiles using semantic search (natural language search), then autonomously calls and interviews them.
Try it yourself: Call +1 (256) 915-2911 and experience an AI interview firsthand. The agent will ask you screening questions and adapt based on your responses. Takes ~5 minutes.
Full platform handles sourcing → calling → interviewing → ranking candidates.
Tech stack: - Gemini LLM for conversational intelligence and real-time response generation - Custom speech-to-text + text-to-speech pipeline for natural voice interaction - WebRTC for low-latency audio streaming (open-source components) - ML models for keyword extraction, skill scoring, and behavioral analysis - Real-time orchestration layer that manages interview flow and timing
The interesting challenge is making the conversation feel natural at scale. We're handling multi-turn interviews with dynamic follow-ups, interruption handling, and context retention across the entire conversation.
Current limitations: - Still struggles with noisy environments - Latency around 800ms could be better - Limitation scoring for soft skills
Architecture questions we are curious about: - How would you reduce latency further? We're at ~800ms response time - Better approaches for handling conversational interruptions? - Thoughts on evaluating "soft skills" algorithmically?
Call the number and let me know where the AI breaks down or if you'd like to know more.
I'm building an AI recruitment platform and your comments really resonated with me - they really show why the current approach is broken.
We actually created this mess by making applications too easy - technology removed friction in the hiring process and now people click apply to jobs without even reading the JD.
Worst still with genAI people that may not be a good fit for the roles they apply, use AI to tweak their CVs to match the job descriptions and become a good fit on paper.
The results is that companies get flooded with irrelevant applications and respond with more friction, whether that is more screening rounds, more interviewing or AI interviews that add the friction back.
We're trying to solve this by adding good friction instead of bad friction. The AI should be working for candidates too, not just extracting value from them.
Would love your thoughts on what would actually make an AI interview worthwhile from a candidate perspective.