mukulmunjal

Hi HN,

We recently crossed our first 100 active users at InterviewFlowAI. Getting here was significantly harder than I originally expected. Looking back, I wasted months of engineering and marketing effort going down the wrong paths.

If I were starting over today, here is the technical and business reality check I wish I had:

Traction over pitch decks: We spent a lot of time prepping for a $1.5M seed round before we had real, consistent usage. Validate your idea by fighting for those first 100 users first. Real traction flips the dynamic of investor meetings entirely.

Stop over-engineering the landing page: As technical founders, we want to build everything custom. Do not build your marketing site from scratch. Study the structure of SaaS tools you admire, adapt their section layouts, and move on to the actual product.

Distribution requires daily work: Writing code is only half the battle. You have to publish online daily. Pick your channel (Reddit, X, LinkedIn, or an SEO blog) and stay consistent to get in front of your target audience.

Give away the hard stuff first: Manual resume screening is fundamentally broken. We built the objective AI engine we wished we had and kept it completely free for a limited time. That raw usage gave us the edge cases, feedback, and testimonials we needed to confidently introduce our $299 and $499 monthly enterprise plans.

The first five minutes dictate churn: Your onboarding is a promise. We realized we had to put a massive amount of effort into those initial screens so hiring managers could experience the AI's value immediately.

Ship smaller pieces faster: Maintain an MVP mindset for every feature. We pushed out tools like our "smart screening questions" quickly just to see how users interacted with them, then used that telemetry to improve the logic.

Incumbents are a target, not a wall: Just because massive legacy ATS platforms exist does not mean you cannot compete. They are often slow and bloated. Focus strictly on execution. We simply aimed to build a better, unbiased scoring engine.

Co-founder alignment is your biggest moat: Having a technical co-founder like Tushar who matches your ambition is the single greatest advantage. The emotional and technical workload is too heavy to carry alone.

Do things that do not scale: Earning those first paying customers is an absolute grind. You have to do the manual, unscalable work to get them onboarded.

Building a solid product ultimately comes down to ignoring your own assumptions and obsessing over what your users actually need.

Happy to answer any questions about our stack, how we transitioned to paid tiers, or the resume parsing logic.

Hey everyone, Mukul here. To add a bit of technical context, the biggest headache we are actively navigating right now is balancing the audio latency with the time it takes the LLM to process the candidate's last answer and generate a relevant, probing follow-up.

We are currently leveraging Supabase and AWS to handle the backend routing and logic. Keeping the conversation feeling natural without awkward pauses is our absolute top priority. If anyone has experience shaving off milliseconds in real-time conversational AI setups, I am all ears.

Also, I know automated interviews can be a controversial topic when it comes to candidate experience. We really tried to make this feel like a helpful screening chat rather than a robotic interrogation. If you decide to test it out, I would love your brutal honesty on whether we are hitting that mark or totally missing it.

Happy to answer any questions!

I’m building an AI tool that conducts first-round interviews automatically to reduce manual screening for hiring teams.

Before taking this further, I’d really appreciate honest feedback on a short demo video of the product.

Things I’m hoping to learn: • Is the problem and value clear? • Does the solution make sense? • Anything confusing or missing?

Demo video: https://www.tella.tv/video/interview-flow-ai-automating-cand...

Thanks in advance to anyone who takes the time to share feedback.

Totally fair concern, and I agree that "automate it away and forget about it" is a fast path to a team full of posers.

Couple of clarifications on how we use this: • It is not “modern looking = good." A repo with years of steady commits on an "ugly” legacy app can score very high if there is depth, complexity, and real usage. • We explicitly look for things like long term maintenance, non trivial diffs, real world integrations, and documentation, not just shiny landing pages. • The output is a signal for a human, not a gate that replaces one. It helps a recruiter who sees 200 links decide which 10 are worth a closer look.

In other words, it is meant to reduce noise, not replace judgment. If someone is quietly maintaining a critical FOSS project, they should stand out more, not less

We noticed something odd while reviewing thousands of candidates: resumes were giving us an increasingly distorted picture of actual ability. People can list anything. Their real work is almost always sitting in a GitHub repo, personal site, Figma file, or some half-forgotten project link.

So we built portfolio-level screening into InterviewFlowAI.

When a candidate submits a portfolio URL, we analyze the actual project: – structure and depth of work – tech stack used (and whether it looks “real”) – clarity of writing and documentation – attention to detail – common red flags (broken links, empty repos, template sites, shallow descriptions)

From that, the system generates: – a match score – strengths and gaps – a short summary the recruiter can skim in a few seconds

Early tests show that portfolio signals correlate far more with real-world ability than anything in a resume, especially for engineers and designers.

Curious if others here have seen the same shift: Are resumes becoming noise? Would love thoughts from people hiring at scale or candidates who prefer showcasing actual projects.

I get the concern. The goal isn’t to replace real human conversations. It’s to remove the repetitive early screens where both sides usually gain nothing. The idea is to let recruiters spend more time on the real human interaction that actually matters, not less.