The Grand Journal of
Project കുമിള (Kumila) ☕

Before we drag you into the midnight panic, the existential dread of having zero code, and our obsession with tea foam, we realized something criminal: we completely forgot to introduce ourselves!

Allow us to introduce the two masterminds (and by masterminds, we mean two sleep-deprived MCA students with caffeine in our blood and an unhinged dream) representing Team realFighters:

🥋 Chief Panic Officer

Zen Varghese

1st Year MCA • Co-Founder

"The guy who stared at the clock at 11:00 PM, mourned our SIH heartbreak, remembered the legendary Appam bubble myth, and dragged Midhun into this glorious mess."

☕ Chief Chaya Officer

Midhun Mathew

1st Year MCA • Co-Founder

"The calm strategist who evaluated Idiyappam rings and mixture peanuts at 11:53 PM, took a slow sip of tea, and dropped the immortal decree: 'Chaya bubble edukkam da!'"

To understand how we ended up building software that counts foam bubbles in tea, you have to understand our tragic, desolate origin. We are first-year MCA students. Before this, during our undergraduate college years, we lived in a virtual technological desert. Our previous college didn't have any tech communities, zero TinkerHub chapters, zero dev clubs, and nobody ever hosted a single hackathon. Not even a mini one.

So when we entered our first year of MCA, we made a solemn vow: we are going to compete in a hackathon no matter what. The golden ticket appeared almost immediately — the prestigious Internal Smart India Hackathon (SIH). We were hyped. We put our game faces on. We started hunting for teammates to conquer the nation.

And that is when disaster struck.

We went around looking for team members. What did we get? Absolute, deafening silence. Nobody was interested. Nobody wanted to form a team. Nobody had any good ideas, and the people we asked vanished faster than free samosas at an orientation. Because we couldn't assemble a team or lock in an idea, we didn’t even get to participate in internal SIH.

We were crushed, rejected, and left standing empty-handed. But surrender is not in our blood. Instead of sulking, I (Zen) turned to my friend Midhun with a wild proposal: "Bro... if the world doesn't want our useful ideas, let's join USELESS PROJECTS 3.0!"

With 60 minutes ticking on the clock, we had no team name, no project idea, and no plan. Under peak deadline panic, we registered ourselves with the most aggressively ironic, cinematic name we could conjure: Team realFighters. We weren't fighting bugs or world hunger — we were fighting the countdown timer!

Now came the hardest challenge: What on earth do we build? In sheer desperation, we opened ChatGPT and Claude. We begged the AI for ridiculously useless ideas. But the AI models failed us completely — they kept generating boring, polite, corporate "anti-productivity" apps that still smelled like homework. None of them had soul. None of them made us laugh.

That’s when a legendary memory flashed in my mind. I recalled a legendary project from a past hackathon: The Appam Bubble Detector — an invention so absurd and so iconic that it became folklore.

Using the sacred Appam Bubble lore as our guiding star, our midnight brainstorm unlocked The Holy Trinity of Kerala Snack Ideas:

We had both options on the table. Zen looked at Midhun. Midhun stared at the screen with the clarity of a monk at a tea stall at 6 AM, and gave the final, immortal verdict:

And so, we took the Chaya Bubble. Form submitted. Registration confirmed. Team realFighters was officially registered for Useless Projects 3.0 with just 7 minutes to spare. (At that frantic midnight hour, we didn't even have an official name for the project yet — the name കുമിള was decided much later, after ending the project in the hackathon!). Who needs Smart India Hackathon when you have the froth of Kerala tea running through your veins?

Everything was smooth sailing in our heads. We had registered for Useless Projects 3.0, we had our glorious mission—Project കുമിള—and the roles in Team realFighters were crystal clear from minute one:

Zen Varghese

Backend & AI / CV Engineer — Computer vision, segmentation models, FastAPI + OpenCV backend.

Midhun Mathew

UI Mastermind & Future CEO — Frontend design, product vision, and the man who will run the company someday.

"Bro, how hard could it be to count bubbles in a glass of tea?" we asked ourselves. Famous last words.

The moment we actually looked at high-resolution photos of fresh Kerala meter-chaya, reality slapped us in the face. Tea froth isn't neat circles on a whiteboard. It’s an organic, chaotic storm of overlapping membranes, microscopic foam, light reflections, caramel shadows, and translucent semi-invisible bubbles! We literally had no clue how to detect them.

The AI Model Gauntlet (Or: How Deep Learning Gave Up on Kerala Tea)

I (Zen) rolled up my sleeves with FastAPI and OpenCV as the backend backbone, and unleashed every cutting-edge AI and segmentation model I could get my hands on. Here is what happened:

🔬 "Behold: The specimen scan results that made us question our life choices..."

← INSPECT ANOTHER ചായ

☕ ENABLE DEBUG MODE

ABCDEF

12345

[ 0 ]

r: px   conf: 0%

ANALYZED SPECIMEN • MULTI-SCALE CV

FILTER BY SIZE: All (74) Small (48) Med (19) Large (7)

TOTAL ചായ BUBBLE POPULATION

74

BUBBLES DETECTED

• MULTI-SCALE CENSUS COMPLETE • HIGH-FIDELITY OPENCV

SPECIMEN CLASSIFICATION

Organic Population

48

SMALL (≤12PX)

19

MEDIUM (13-24PX)

7

LARGE (≥25PX)

📸

Archival Note regarding this Specimen View:
The interactive interface above is reconstructed directly inside our new UI to showcase the flawed detection results (it tagged only 74 bubbles, completely missing hundreds of real, fine foam bubbles beneath the surface!). Our very original, messy early-experiment screenshots were lost to the digital void during our late-night panic, so we memorialized the exact flawed data in this new interactive view.

The Honest Hacker's Blunder: "Wait... You Mean We Didn't Have to Delete It?!"

While experimenting with these crazy detection techniques, the official kickoff time for Useless Projects 3.0 was approaching.

I sat back, proud of my ironclad integrity. "Now we will code like true realFighters from minute zero!"

And then the hackathon started. We logged into the channels and saw what the other participants were doing: EVERY SINGLE TEAM HAD ALREADY BUILT HALF OR ALL OF THEIR PROJECTS! People were showing off polished repos, pre-trained pipelines, and working prototypes! Meanwhile, Team realFighters were sitting there with a freshly wiped directory and an empty hard drive.

Midhun to the Rescue... Or So We Thought! ☕📉

As sheer terror washed over me, Midhun leaned back with the calm swagger of a Fortune 500 CEO and smiled:

My heart leaped! We were saved! The future CEO had clutched the round! I scrambled over to check his machine, expecting a miracle. Instead, I stared into the abyss of Pure Vibe Coding™.

Midhun had decided to bring Computer Vision into the client-side browser using pure JavaScript and OpenCV.js. The catch? He had vibe-coded the entire thing without knowing what algorithms, matrix filters, or thresholding mechanisms were actually happening under the hood!

The browser tab choked, froze, and died. OpenCV.js was getting stuck in infinite loops, the detection was failing catastrophically, and neither the browser nor Midhun knew what the math was doing. At that exact moment, at 2:00 in the morning, with all my clean backend deleted and Midhun's vibe-coded JS melted, I thought: We are 100% doomed.

The Executive Decree: "Start From 1!" 🚀

You can't rattle a future CEO for long. Midhun closed the frozen browser tab, took a deep breath, looked at our cold mugs of tea, and laid down the law:

With that decision locked in, the real fight began. We wiped the sweat from our foreheads, brewed another round of strong tea, and got to work building the true, immortal Project കുമിള.

In the middle of our existential crisis, something magical happened. We glanced to our right at the hardware team sitting next to us — the ones building actual circuits and soldering actual components like actual engineers. We figured, why not break the awkward silence and introduce ourselves?

What are the odds?! Our project — the idea we panic-registered at 11:53 PM — was apparently a past-year Useless Projects 2.0 legendary project! The universe was telling us something. Unfortunately for the hardware team, they never fully completed or presented it at the end. Their unfinished dream had now become our very much alive (and barely functioning) reality.

But the most important thing came next. We asked them — how did you approach bubble detection? And they said two words that changed everything:

That was it. That was the key. I (Zen) immediately pulled up ChatGPT, typed in "circle detection for bubble counting OpenCV" and ChatGPT, for the first time in this entire journey, actually delivered. It told us about contour detection + HoughCircles — classic OpenCV algorithms perfectly suited for detecting circular bubble shapes in images.

We implemented it. We ran it. We held our breath. And then...

📸 ACTUAL DETECTION OUTPUT • MULTI-SCALE CV CONTOUR ANALYSIS

There she is. 74 bubble circles detected. Classified by size. Marked with orange rings. Was it a perfect result? No. Were there hundreds more invisible micro-bubbles we completely missed? Absolutely. But for a project born at midnight, running on zero prior experience and a ChatGPT tip from a hardware bro, this was a certified miracle. 🎉

While I (Zen) was wrestling with OpenCV circles and contour detection in the backend, Midhun took on what turned out to be the hardest part of the entire project: the frontend UI.

And Midhun being Midhun — he did not just slap together a basic HTML page. No. He went on a research expedition. He dived deep into Canva, Stitch, Behance, Dribbble, and every design inspiration website known to mankind. He studied aesthetic palettes, hand-drawn scrapbook vibes, journal-style layouts, warm paper textures, and typographic hierarchies that would make a design school professor weep with joy.

The result of all this research? A stunning, warm-toned, handcrafted scrapbook interface that made our useless bubble detector look like a premium scientific instrument. And somewhere in that beautiful chaos, Midhun decided to include a very special guest:

🇮🇳 SPECIAL GUEST APPEARANCE

Modi Ji Enters the Chat

Yes. The Prime Minister of India made a cameo appearance in our chaya bubble detector UI. Midhun, in a stroke of absolute unhinged genius, added a tiny Modi ji somewhere in the interface — complete with a thumbs up — as the ultimate symbol of government-approved bubble science.

Was this planned? No. Was this necessary? Absolutely not. Did it make the project 1000% better? Without question.

🏛️ "Official bubble detection, verified by the highest office in the land."

The UI research, the Canva inspiration boards, the Stitch explorations, the hand-picked Google Fonts, the warm paper aesthetic, the scrapbook-journal feel — all of it came together into something that was genuinely beautiful. Which created a new problem: our UI was now better than our actual project.

Here is where things went from useless to magnificently, absurdly useless. We were already embarrassed by our project. Compared to the other teams — whose projects were funny, creative, impressive — ours felt like a negative funny useless project. The bubble detector wasn't getting laughs. We needed more.

The idea: ChaiScript — a fake programming language where you code a cup of tea step-by-step using terminal commands, and the system compiles your recipe into a virtual brew with a bubble profile analysis. Absolutely useless. Absolutely perfect.

The execution: Midhun, through Antigravity vibe coding, built the entire ChaiScript terminal compiler in the frontend. In what can only be described as "coding by vibes and prayer," the terminal came to life:

01

THE TERMINAL IS BORN

Behold — CHAI_TERMINAL v1.0, powered by the ChaiScript Compiler. Type chai.init to begin your culinary programming journey. Add tea, milk, sugar. Type chai.brew. Watch science happen. This is simultaneously the most pointless and most satisfying thing we built.

Verdict: It runs. It compiles. It outputs tea instructions. Zero actual beverages produced. 100% satisfaction guaranteed.

But wait — what happens after you submit the recipe? You can't just have a terminal that does nothing. The ChaiScript compiler needed to actually compile your chai. And so, the compilation screen was born:

02

COMPILING CHAI RECIPE...

"Parsing recipe specification... Recipe syntax valid. Loading tea extraction module... Tannins & color profile locked. Loading dairy emulsion module... Milk fat balance calibrated. Heating & Aerating Chai... 56%..."

At this point during the presentation, at least three audience members genuinely leaned forward to see if the tea was actually going to be ready. It was not. The progress bar is a lie. It will always be a lie. That is the art.

Verdict: The most realistic fake compiler in Kerala. "Tannins & color profile locked" is a sentence that has never been spoken in software engineering. Until now.

And finally — after all the compilation, all the chai.init commands, all the tannin calibrations — you get your result. The moment of truth. The masterpiece:

03

COMPILATION COMPLETE — YOUR CHAI IS READY

The ChaiScript compiler presents your final brew: a rendered cup of tea with specs. Foam Density: 50%. Bubble Profile: MEDIUM. Tea Strength: 100%. Boil Intensity: 50%.

It even says at the bottom, in small italic text: "100% scientifically redundant."

Which honestly describes the entire project. And we are incredibly proud of that.

Final verdict from our own UI: "100% scientifically redundant." We couldn't have said it better ourselves. Also — there is a button that says "COMPARE WITH REAL ചായ". We did not implement that button. It does nothing. Perfect. 10/10.

The day of reckoning arrived. Useless Projects 3.0 presentation time. We had submitted the project. The code was running (mostly). The UI looked beautiful. Modi ji was in position. Everything was technically ready.

What followed was a presentation masterclass in how things can go beautifully, spectacularly wrong.

After the presentation ended, we walked away from the stage with the specific kind of silence that only two people who just flopped together can share. No words were needed. We had built something, submitted it, presented it, and watched it land with the grace of a wet chapati on a marble floor.

In the aftermath of the spectacular flop, as we sat in existential contemplation of our bubble-counting legacy, something wonderful happened. Something that made all the midnight panic, the deleted code, the frozen demos, the invisible Modi ji, the trembling hands — all of it — completely worth it.

And that is the thing about Team realFighters. We didn't win. We didn't impress anyone. Modi ji wasn't noticed. The demo froze. The hands trembled. The presentation flopped. But we showed up, built something from nothing, submitted it before the deadline, and stood on that stage.

First-year MCA students. Zero prior hackathon experience. Registered at midnight. Built in chaos. Presented with trembling hands. And at the end, we got stickers and vada.