Mygom.tech

Text Surveys Are Where Honest Feedback Goes to Die

"Rate your experience 1-5." "Please elaborate." Nobody types three paragraphs. You get vague answers. "It was fine." "Could be better." Nothing actionable.

Video captures real reactions, genuine emotion, nuances text can't. But now you're watching hours of footage, manually transcribing, spotting patterns. Still slow.

Our client wanted video feedback without the headache.

The Challenge:

Build a platform where anyone creates video surveys, collects authentic responses, and gets instant AI-powered insights—without manually reviewing hours of footage. Support all devices, handle multiple user roles, provide real-time analytics, make it simple.

What We Built:

Easy Survey Creation: Customizable templates with conditional logic. Interactive surveys that adapt based on responses.

Smooth Recording: Record on any device. Pause, resume, re-record until satisfied. Higher completion rates.

AI Analysis: Automatic transcription, emotional tone detection, sentiment analysis, summaries. Ask in plain language: "What were the main concerns?" Get instant answers.

Real-Time Dashboard: Live updates as responses arrive. Track themes, completion rates, progress.

Tech Stack: Next.js, NestJS, PostgreSQL, Daily.co, OpenAI, Stripe, Cloudflare R2

The Results:

• 80% accuracy in emotional tone detection
• 70% precision in sentiment classification
• 75% accuracy in overall video analysis
• Instant insights replacing hours of manual review
• Higher response rates from better UX

The Impact:

Market research that meant watching hours of video and guessing at sentiment—now delivers AI-analyzed insights the moment responses arrive.

Video feedback that's actually usable.

Buying 10,000 Kilos of Coffee Shouldn't Require 47 Emails and Three Weeks

Restaurant chain needs bulk ingredients. Emails 12 suppliers. Waits for quotes. Follows up. Gets scattered responses in different formats. Can't compare. Negotiations drag on for weeks. Deal closes—or falls apart.

Bulk food trading was stuck in the 1990s. Our client wanted to fix it.

The Problem:

Buyers sent RFQs via email, got scattered responses, couldn't verify supplier credibility, and spent weeks negotiating manually. Suppliers competed blind on pricing with limited buyer visibility. No transparency. Just slow procurement cycles that killed deals.

What We Built:

A modern B2B marketplace where bulk food trading actually works:

Structured RFQs: Buyers post requests. Multiple verified suppliers submit competitive bids. Compare offers side-by-side in one place.

Smart Catalog: Filterable listings with specs, certifications, quality data. Find what you need based on real requirements.

Real-Time Communication: Built-in messaging and notifications. Questions answered immediately. Negotiations happen fast.

Verified Profiles: Both sides see credentials upfront. Trust built before the first message.

Tech Stack: Next.js, NestJS, TypeScript, PostgreSQL, TypeSense, Ably Chat, Google Cloud

The Results:

• 60% faster procurement—weeks reduced to days
• 3x more vendor-buyer matches
• 100% transparency with verified credentials
• Higher conversion rates from clearer negotiations

The Impact:

What took 47 emails and three weeks now happens in days on one platform. Buyers find suppliers confidently. Suppliers reach new markets. Everyone closes more deals.

Bulk food trading that finally feels modern.

The Problem:

Invoices arrived as emails, PDFs, images, spreadsheets. Finance teams re-typed data, matched payments manually, fixed errors by hand. Result? Duplicate payments costing thousands, missed discounts, forgotten subscriptions, and weeks-long audits.

We had this problem at Mygom. So we built a solution for ourselves, proved it works, and now deploy it for other companies.

What It Does:

AI Capture – Automatically collects and reads invoices from email and folders (PDFs, images, spreadsheets). Extracts vendor, amount, dates, invoice numbers. No manual entry.

Smart Reconciliation – Matches invoices to bank payments with 95% accuracy, even when descriptions don't match. No more manual reconciliation.

Financial Protection – Flags duplicate payments before they go out. Detects early-payment discounts. Tracks subscriptions and finds unused seats or forgotten renewals.

Clear Dashboard – Real-time view of all invoices, payments, and subscription spend. Audit-ready records.

Tech Stack: Next.js, NestJS, PostgreSQL, TypeScript, Python, OpenAI

Results:

• 40% faster invoice processing
• $2,000+ saved per prevented duplicate payment
• Immediate detection of forgotten subscriptions
• Zero missed early-payment discounts
• Hours reclaimed from manual work

We built this because we needed it. Other finance teams use it to cut costs, prevent errors, and automate invoice chaos.

Finding Office Space Shouldn't Feel Like Online Dating With Your Eyes Closed

Business needs office space. Searches for weeks. Gets irrelevant listings. Contacts landlords. Waits for responses. Landlord wonders: can they actually pay? Tenant wonders: is this landlord legit? Weeks turn into months. Deals stall. Everyone's frustrated.

Commercial real estate leasing was broken. Our client wanted to fix it.

The Problem:

Tenants drowned in irrelevant listings—spaces that looked fine until you realized there's no loading dock or the ceiling height won't work. Landlords couldn't verify if tenants were financially stable. Brokers sat in the middle, adding fees and delays. Basic search filters weren't enough. Nobody trusted anybody.

What We Built:

An AI-powered platform that actually understands what matters in commercial real estate:

Smart Matching Engine: Goes beyond "square footage + location." Analyzes detailed requirements—ceiling height, loading dock access, proximity to highways, parking capacity. AI learns from user behavior and gets smarter with every search, delivering genuinely relevant matches.

Financial Transparency: Integrated Proff.no API gives landlords instant access to tenant financial profiles in real-time. Tenants see landlord histories. No more guessing. No more surprises. Just clear data for both sides.

Direct Communication: Built-in messaging. No brokers. No middlemen fees. Tenants and landlords talk directly. Compatibility scores highlight best matches immediately.

Modern Stack: NextJS, Tailwind, NestJS, Prisma, PostgreSQL. Fast, reliable, easy to maintain. ChatGPT-powered assistance and Propcloud data integration keep information current.

The Results:

• 40% faster property matching
• 2x quicker tenant-landlord connections
• 100% transparency on both sides
• Lower costs by eliminating broker fees

The Impact:

What used to take months of searching irrelevant listings and waiting for broker responses now happens in days—with confidence that the match actually makes sense and both parties are credible.

Commercial real estate leasing that finally works like it should.

What If an Insurance Claim Took 10 Minutes Instead of 10 Days?

Hail hits. Thousands of cars damaged. Insurance adjusters drive to each one, count dents by hand, write reports. Claims take days or weeks. Customers wait. Insurers pay for manual labor.

Our client built AI that could detect and count hail damage from video. Proven in the lab. But before rolling it out to insurers, they needed proof: does it work in the real world with regular phones?

The Challenge:

Test the entire workflow with real cars, everyday phones (iOS and Android), and actual field conditions. Mobile video to automated damage report in under 10 minutes. No manual intervention. Prove it scales.

What We Built:

A lightweight web app that turns any smartphone into an AI damage assessment tool:

Smart Video Capture: Guided filming of specific vehicle areas (hood, roof, sides) with built-in visual instructions. Ensured consistent footage quality across devices—critical for AI accuracy.

Instant Processing: Videos uploaded automatically. AWS Step Functions triggered the backend: video reformatting → AI dent detection → damage analysis → PDF report. Fully automated.

Real-Time Reports: Detailed PDF with dent counts and damage assessment, delivered within minutes.

Built for Scale: Logged every detail across 600+ test vehicles. Multi-device support proved cross-platform reliability.

Tech Stack: Next.js, NestJS, PostgreSQL, AWS (EC2, Lambda, SES), Python, React Native

The Results:

• 500+ vehicles processed in real-world conditions
• 80%+ reports generated in under 10 minutes
• 5x faster than traditional inspections
• Zero manual counting required

The Impact:

What once required scheduling adjusters, driving to locations, manually counting dents, and writing reports over days—now happens in minutes from a phone video.

Insurance claims that took weeks now take 10 minutes.