All blogs · Written by Ajitesh
How Bambinos Scaled Hiring 7x with Voice AI (and Built It in 4 Days)
A conversation with Krishna Prasad on transforming recruitment at a Shark Tank-backed edtech startup
Introduction
What if you could build an AI hiring system in four days—without writing a single line of code—and scale your recruitment from 10 to 70 hires per month?
That’s exactly what Krishna Prasad, Product Manager at Bambinos, accomplished. And the results speak for themselves: 80% time saved on screening, 24-hour application-to-onboarding cycles, and higher quality candidates.
I sat down with Krishna to understand how he took Tough Tongue AI from discovery to production in weeks, what challenges they faced introducing AI to traditional hiring processes, and why some candidates now prefer AI interviews over human ones.
This isn’t a theoretical case study. It’s a real story about a non-technical PM who shipped an AI product fast, tested it with 5,000 users, and transformed how an entire company hires.
The Challenge: Hiring Doesn’t Scale
Bambinos is a Shark Tank-funded edtech startup revolutionizing how children ages 4-14 learn. They offer courses in English, Math, Geeta (cultural heritage), and other subjects—all delivered through innovative, AI-enhanced teaching methods.
As they grew, hiring became a bottleneck.
”Earlier we were hiring only 5 to 10 teachers per month,” Krishna explained. “Our talent acquisition specialist had to make hundreds of calls a day just to know if the candidate is really available or okay for this role.”
The math was brutal. For every teacher they needed to hire:
- 100+ screening calls to find interested candidates
- Manual resume review for every application
- Inconsistent quality across hires
- Weeks of back-and-forth scheduling
And the problem wasn’t unique to teachers. They needed to hire for sales roles, customer success, operations—across multiple functions.
Traditional hiring processes simply don’t scale. You can throw more recruiters at the problem, but that’s expensive and still doesn’t solve the consistency issue. Different interviewers ask different questions, evaluate candidates differently, and have different availability.
Krishna knew there had to be a better way.
The Discovery: Finding Voice AI That Actually Works
Three months ago, Krishna attended an event where a friend was using Tough Tongue AI.
”I was curious,” he told me. “When there are so many products in the market like Vapi or ElevenLabs, what differentiates Tough Tongue?”
As a product manager from a non-technical background, Krishna had tried building with other voice AI tools before. The experience was frustrating. Setting up Vapi or ElevenLabs required significant technical knowledge, and he needed to ship fast.
”I focus on building a product in less than 10 days and trying it out. That’s my policy. Bring it to market.”
What caught his attention about Tough Tongue AI was how quickly he could get started. No complex setup. No wrestling with APIs and webhooks. Just plug and play.
”I built a prototype in Lovable and added Tough Tongue’s API. We were able to launch the product in less than four days.”
Wait—four days? From idea to working product?
”Yes, and myself, I was able to make sure that my prototype built without even a single line of code was able to handle 5,000 users.”
The Build: From Prototype to Production
Krishna’s approach was methodical but fast.
Phase 1: Initial Prototype (Days 1-4)
Originally, Krishna wasn’t even thinking about hiring. His first use case was training their sales team.
”On the initial phase, I had built it for training our salespeople. Sales agents need to practice their pitch, they need to self-validate.”
But then he noticed something. Their talent acquisition team was drowning in screening calls. One person was making 100+ calls daily, trying to get candidates to come to the office for interviews.
”It’s better—why can’t we implement it in a unique way where no one ever tried it?”
So he pivoted. Using Lovable (a no-code development platform) and Tough Tongue AI’s API, he built a hiring interview system. The AI would conduct the first-round interview, asking candidates questions, evaluating their responses, and making pass/fail decisions.
”First thing was a failure,” Krishna admitted. “Then we built another model. That worked well. It was built in four days.”
Phase 2: Testing at Scale (Month 1)
Here’s where most companies would have stopped. Build a prototype, show it to stakeholders, schedule meetings to discuss rollout plans, form a committee…
Not Krishna.
”We launched it and we onboarded around 5,000 users. It was easy for us to evaluate Tough Tongue’s capacity.”
Five thousand users. On a prototype. Built without traditional development.
This is where modern product management gets interesting. Krishna wasn’t asking for permission. He was testing with real users, gathering real data, and proving the concept worked at scale.
Phase 3: Productionization (Month 2-3)
Once they validated that the prototype could handle volume, they transitioned to traditional development for long-term scalability.
”We had a transition to traditional development for scaling purposes. On the initial phase, it was difficult. Bringing AI for hiring, a voice model—AI agent is going to take an interview—how do people feel?”
The complaints flooded in. Candidates weren’t sure what to expect. Some were uncomfortable talking to AI. Others experienced technical issues with audio or video.
”Our groups were fledged with complaints. But with the help of Ajitesh and his team, we were able to sort it out day by day. We used to sit midnight or 24/7. We used to work together.”
This is the reality of shipping AI products. The technology works, but the human elements—expectations, experiences, comfort levels—need iteration.
How It Works: Inside Bambinos AI Hiring System
During our conversation, Krishna showed me the system in action. It’s impressively seamless.
The Candidate Experience
- Application: Candidate uploads their resume on Bambinos website
- Auto-parsing: The system automatically extracts experience, skills, and relevant details
- AI Interview: The candidate immediately starts a voice/video interview with the AI
- Hyper-personalization: The AI references specific details from their resume, creating a tailored conversation
- Evaluation: Within 5-10 minutes of completing the interview, the candidate receives their result
The entire process happens in their custom-built application. No third-party dashboards. No redirects. It’s completely embedded in Bambinos hiring flow.
What Makes It Different
Most AI hiring tools analyze transcripts. They convert speech to text, evaluate the words, and make decisions based on what was said.
Bambinos system is multimodal. It analyzes:
- Voice modulation and tone: Is the candidate confident? Hesitant? Enthusiastic?
- Facial expressions: Do they maintain eye contact? How do they react to questions?
- Body language: What does their posture and presence communicate?
- Content quality: Are their answers substantive and relevant?
”Unlike text-based screening,” Krishna explained, “we’re evaluating everything a human interviewer would notice.”
The AI asks rigorous questions—the same questions a founder or HR manager would ask. It probes on experience, tests communication skills, and evaluates role-specific competencies.
”The candidates we get after the filtration of AI are really amazing,” Krishna said. “The person who clears the AI interview is really amazing because our AI interview is not that much easy.”
The Results: 7x Growth and 80% Time Saved
The numbers tell the story:
Hiring Volume
- Before: 5-10 teacher hires per month
- After: 50-70 teacher hires per month
- Growth: 7x increase
Screening Efficiency
- Before: 100+ manual screening calls daily
- After: AI handles all first-round screening
- Time saved: ~80% of talent acquisition time
Candidate Experience
- Before: Days or weeks to schedule and complete first interview
- After: Immediate interview upon application, results in 10 minutes
- New promise: “24-hour hiring policy”—application to onboarding in one day
Quality of Hires
This is the surprising part. You might expect that automating screening would reduce quality. The opposite happened.
”Around 80% of our time is saved, and the impact and distribution is also really working for us,” Krishna shared.
Why? Because the AI consistently applies the same rigorous evaluation criteria to every candidate. No interviewer fatigue. No bad days. No unconscious bias based on how someone looks or where they went to school.
Every candidate gets asked the right questions, evaluated on the right criteria, and judged on their actual performance.
The Surprising Truth: Candidates Prefer AI Interviews
This was the most counterintuitive finding.
”Few of our candidates texted me that it was pretty easier than talking to a human,” Krishna told me.
Wait—easier? How could talking to an AI be easier than talking to a person?
”Because they didn’t have hesitation to ask questions to HR. AI doesn’t have that kind of thoughts, so it was easy for asking questions.”
Think about that. In a human interview, candidates often hold back. They don’t want to seem uninformed. They don’t want to ask “stupid questions.” They’re worried about making a bad impression.
With AI, that anxiety disappears. The AI isn’t judging them for clarifying a question. It’s not getting impatient if they need a moment to think. It’s just there to evaluate their actual competence.
Krishna was quick to acknowledge not everyone feels this way. “So many candidates said we are not comfortable with AI. It is implementing a new thing. We will be having both sides, right? A coin has two sides.”
But the trend is clear. As people experience well-designed AI interviews, many prefer them to traditional screening calls.
Overcoming Resistance: Getting Stakeholder Buy-In
Introducing AI to traditional processes isn’t just a technical challenge. It’s a people challenge.
”When I approached my stakeholder, my founder, like we have a new solution—why can’t we implement it? He’s like, okay, let’s try it out.”
Krishna got lucky with an open-minded founder. But even with approval, he faced resistance from:
Candidates: “Why are you making us talk to AI?” HR team: “How do we know this actually works?” Sales leadership: “Will this work for roles beyond teachers?”
His approach was pragmatic: show, don’t tell.
”What I did is I simply had free minutes from Ajitesh for testing out. So I built it, tested it, and showed the results.”
Instead of PowerPoint decks and lengthy proposals, he built a working prototype, ran real interviews, and presented the data.
This is the modern PM playbook:
- Get scrappy access to tools (free trials, pilot programs)
- Build fast with no-code/low-code tools
- Test with real users
- Let results speak for themselves
”We made a motto: 24-hour hiring policy. Application to onboarding in 24 hours if you are selected. And that really made a very good impact for us.”
Once they could articulate that promise—hire someone in 24 hours—the value became undeniable.
Expanding Beyond Teachers: Sales, Operations, and More
Success breeds expansion.
”Early we were using hiring for teachers. Now we started implementing it for sales roles and other multiple roles.”
Each role requires different evaluation criteria. For teachers, they assess subject knowledge, communication clarity, and patience. For sales roles, they look at persuasion skills, objection handling, and energy.
The beauty of Tough Tongue AI’s platform is that Krishna can customize scenarios for each role without rebuilding the entire system.
”We just define what the AI should focus on for each role. The evaluation criteria, the key questions, the skills that matter.”
This flexibility means they’re now using AI screening across their organization:
- Teachers: Subject expertise, teaching style, communication
- Sales: Pitch quality, objection handling, persuasiveness
- Customer Success: Empathy, problem-solving, product knowledge
- Operations: Process thinking, attention to detail, communication
The Role of Vibe Coding: How Non-Technical PMs Ship Fast
One of the most fascinating parts of my conversation with Krishna was learning about his development philosophy.
”I’m from a non-tech background. It is really difficult for me because I focus on building a product in less than 10 days.”
How does a non-technical PM ship production features in 10 days?
The answer: vibe coding.
What Is Vibe Coding?
Vibe coding (also called AI-assisted development) uses tools like Lovable, Cursor, and Windsurf to build applications by describing what you want rather than writing code from scratch.
Krishna’s process:
- Describe the feature in natural language
- The AI generates working code
- Test immediately with real users
- Iterate based on feedback
- Move to traditional development if it proves valuable
”Early we need to spend a good time in creating PRDs, validating that, talking with tech team. Now it is so simple—you have to write a good prompt.”
The Impact on Product Culture
This shift changed how Bambinos entire team works.
”Whether a tech team or design team or me personally, or founders, everyone started using vibe coding. Now we don’t discuss ideas. We just put some prompt, get the output, share that output to the people.”
Even their sales team uses Lovable.
”More than a product manager, sales people use our CRM daily. I have given them access. They’ll just play with it. They don’t ask me—they build it and share the link. ‘We have done this, can you just check this out?‘”
This democratization of development is profound. Sales people who understand customer pain points can now prototype solutions themselves rather than trying to explain their vision to PMs and designers.
Where to Draw the Line
I asked Krishna where he draws the line between vibe coding and traditional development.
”AB testing has really pretty much started becoming very much easy for us. We vibe code every new product or new feature. We test with around 1,000 to 2,000 people. If it really works, we push it into production with traditional coding. Or else we’ll drop it.”
This is the strategy:
- Vibe coding for: Prototypes, experiments, MVPs, quick iterations
- Traditional coding for: Production systems, scale, performance optimization
”Existing developers use Cursor for building products. It is really acting as a junior developer. Sometimes it sucks, but literally we are able to manage that. We are able to bring into production at very good speed.”
The Technical Reality: What Actually Happened Behind the Scenes
While Krishna built the initial prototype in days, productionization revealed real challenges.
Experience Issues
”There were experience issues because people might be in very different networks, so many devices, phones—getting all of these experiences right.”
Voice and video AI is demanding. It requires:
- Low latency (nobody wants to wait 3 seconds for the AI to respond)
- Audio quality (background noise, bad mics, network issues)
- Video reliability (camera permissions, bandwidth)
- Cross-device compatibility (phones, tablets, desktops)
“On the initial phase, we were facing so many issues. ‘This is not working, this is not working.’ Our groups were fledged with complaints.”
The Midnight Debugging Sessions
”With the help of Ajitesh and his team, we were able to sort it out day by day. We used to sit midnight or 24/7.”
This is the reality of shipping AI. The technology is powerful, but integration with real-world systems requires iteration.
Some issues they solved:
- Audio dropouts on poor connections
- Video freezing on mobile devices
- Inconsistent evaluation across different question types
- Candidates getting stuck in the flow
”It was difficult, but other than the implementation stage issues that all developers face, it was really peaceful.”
The Multimodal Challenge
Most AI hiring solutions convert speech to text, analyze the transcript, and make decisions. That’s easier to build but loses crucial information.
Bambinos uses true multimodal AI. It processes audio, video, and text simultaneously. This is technically harder but produces better results.
”Multimodal AI analyzes voice tone, body language, and facial expressions—not just words. That’s what differentiates us.”
This capability came from Tough Tongue AI’s platform, which handles the complexity of multimodal analysis behind the scenes.
What’s Next: The Future of AI Hiring at Bambinos
Krishna has ambitious plans.
”If you ask me, if Tough Tongue AI can implement emotional screening—checking the emotions, triggering a candidate, making them pitch—that kind of emotional analysis might really help us.”
Right now, the system analyzes tone and expression, but Krishna wants deeper emotional intelligence.
”Especially for hiring sales candidates, emotion adds value. How emotionally are they able to convince clients?”
This points to where AI hiring is heading:
- Personality assessment: Beyond skills, understanding cultural fit
- Stress testing: How do candidates perform under pressure?
- Emotional intelligence: Can they read and respond to emotional cues?
- Creativity evaluation: How do they approach novel problems?
Lessons for Product Managers and Founders
If you’re building AI products or transforming traditional processes with AI, here’s what Krishna’s story teaches:
1. Ship Fast, Learn Fast
Don’t wait for perfect. Krishna built a working prototype in 4 days and tested with 5,000 users. Most companies would spend 4 weeks writing PRDs.
”I focus on building a product in less than 10 days and trying it out. That’s my policy.”
2. Use the Right Tools for the Job
Krishna used:
- Lovable: For rapid prototyping
- Tough Tongue AI: For multimodal voice AI capabilities
- Cursor: For production development
Each tool serves a purpose. Don’t try to build everything from scratch.
3. Test with Real Users, Not Focus Groups
5,000 users on a prototype gives you real data. Not opinions in a conference room.
”It was easy for us to evaluate Tough Tongue’s capacity.”
4. Iterate Based on Real Feedback
When complaints flooded in, Krishna didn’t shut down the project. He iterated.
”We were able to sort it out day by day.”
5. Democratize Development
Let non-technical team members build. Your sales team knows customer pain points better than anyone.
”Even our sales people, if they want some feature added, just build it. Everyone started using no-code tools.”
6. Make Bold Promises (When You Can Deliver)
“24-hour hiring policy” is a bold claim. But when you can actually deliver it, it becomes your competitive advantage.
7. Focus on Impact, Not Technology
Krishna doesn’t talk about AI for AI’s sake. He talks about 7x hiring growth and 80% time saved.
”It’s all about making impact.”
The Broader Implication: AI in Hiring Is Here
Bambinos story is part of a larger shift in how companies hire.
Traditional hiring is:
- Slow: Weeks or months from application to offer
- Inconsistent: Different interviewers, different standards
- Expensive: Recruiter time doesn’t scale
- Biased: Unconscious biases affect every human decision
AI hiring done right is:
- Fast: Minutes to hours from application to first-round decision
- Consistent: Same evaluation criteria for every candidate
- Scalable: Handle 10 or 10,000 applicants with same quality
- Fair: Evaluated on performance, not appearance or background
The resistance Krishna faced—“Why are you making us talk to AI?”—is giving way to a new reality. When candidates experience well-designed AI interviews, many prefer them.
”It was pretty easier than talking to a human because they didn’t have hesitation to ask questions.”
This doesn’t mean AI replaces human judgment. The AI handles first-round screening—the time-consuming, repetitive part. The final decision still involves humans.
”After the first round of filtration, we just need to have a final discussion with them. The AI replaces everyone and takes care of all these roles and asks each and every question that a founder or HR would ask.”
This is the future: AI handles volume and consistency, humans handle nuance and final decisions.
Final Thoughts: The 24-Hour Hiring Future
When I asked Krishna what surprised him most about this journey, he said:
“If you ask me six months back, I had zero idea how it will make an impact like this. We have seen 7 to 10x growth in our hiring process, the quality of candidates we get… Around 80% of our time is saved.”
This is what successful AI implementation looks like. Not PowerPoints and pilot programs. Not six-month roadmaps and steering committees.
Build fast. Test with real users. Iterate based on feedback. Scale what works.
Bambinos went from making 100+ screening calls daily to offering 24-hour hiring. From inconsistent quality to rigorous AI evaluation. From 10 hires per month to 70.
And it started with a curious PM who built a prototype in 4 days.
If you’re thinking about AI for hiring, training, or any other process that requires scale and consistency, the lesson is clear: the tools exist. The technology works. The question is whether you’ll take the leap.
About This Conversation
This blog post is based on my conversation with Krishna Prasad, Product Manager at Bambinos, for the Voice AI Podcast. We discussed how they built and scaled AI hiring using Tough Tongue AI, the role of vibe coding in modern product management, and where voice AI is heading.
Listen to the full conversation: https://youtu.be/G0dai75oROY
Want to try AI hiring or training for your team?
- Explore Tough Tongue AI: app.toughtongueai.com
- Book a consultation: cal.com/ajitesh/15min
- Watch platform demo: youtu.be/EzQxlROT-Oo
Discussion Questions
I’d love to hear your thoughts:
-
For PMs and founders: Have you used vibe coding tools like Lovable or Cursor? What’s worked for you?
-
For HR and talent leaders: Would your candidates prefer AI first-round interviews? What concerns would you need to address?
-
For anyone hiring: What’s your biggest bottleneck in recruiting? Is it volume, quality, speed, or something else?
Drop a comment or reach out directly. I’m always curious to hear how teams are thinking about AI in their operations.
Ajitesh Abhishek is the founder of Tough Tongue AI, an agentic AI platform that creates realistic training and assessment environments. Previously at Google, he’s passionate about making AI accessible and useful for real-world business challenges.
Connect on LinkedIn | Twitter | help@getarchieai.com
