BINAYAK
JHA.

I build AI products that real people use — and lead the teams that ship them. A pandemic platform that helped 900+ families. AI that speaks Nepali to 30 million people. An FX model that's 98% accurate.

Feel the work ↓
CS @ F&M ’27×Warburg Pincus Fellow @ Wharton×AWS All Builders Grant
Binayak Jha
est. kathmandu27.71°N → 40.04°N
5 apps in 5 weeks200+ users each98% fx forecast accuracy$20,000 raised900+ families supported84,000+ people reached1 of 34 nationwide — warburg pincus fellow275+ students trainedpeer-reviewed quantum crypto co-authoraws all builders welcome grant ’25

01the work

Don't read the projects.
Feel them.

Six builds, recreated as living exhibits. Everything below moves, types, and updates — click around. Simulations of the real thing, in miniature.

sim // 1Day2026 —
1Day
day 01
?

This could be you

day 01 · waitlist open

Your name. Your work. Your pitch — 24 hours with the whole community watching.

128 in the crowd

the stage resets in

--:--:--

that's your real local midnight

interactive concept · waitlist live
EX.011Day · Founder

1Day

New builders never get seen — reach belongs to whoever already has it

A social stage where every 24 hours oneperson gets the entire community's attention — show your work, sell your thing, be seen. At midnight, the stage resets. In design now, shipping in public. The countdown in the demo is your real midnight.

1
person on stage
24h
then it resets
0
followers required
Product 0→1ReactNetlify
sim // HamroGPT2023
HamroGPT
हाम्रो = ours

— pick a prompt below —

नेपालीमा सोध्नुहोस्…demo
EX.02Incubate Nepal

HamroGPT

30M+ Nepali speakers; AI products that didn't speak to them

An AI suite tailored for Nepali users — conversation, stories and translation in Devanagari, tuned for how people actually phrase things. I led a team of five mentees to ship it.

+40%
engagement
5/5
user satisfaction
5
mentees shipped it
PythonNLPFastAPIReact
sim // Beds & Oxygen, Live2021
covidresources.np
wave II · jun 2021
districticu bedso₂ cylstatus
Kathmandu1241OK
Lalitpur718LOW
Bhaktapur49CRITICAL
Kaski922OK
Chitwan511LOW
Morang815OK
updated 0s ago · 77 districts tracked
$0
raised
0+
families
0+
reached
EX.03Covid Resources Nepal

Beds & Oxygen, Live

Wave II: families dialing hospitals one by one, in the dark

A real-time tracker of hospital beds and oxygen cylinders across Nepal, built mid-crisis by a volunteer team. I led research and the comms that moved relief money where it mattered.

$20K
raised for relief
900+
families supported
84K+
people reached
Web appReal-time dataCampaigns
sim // Forecasting the Rupee2024
USD/NPR· t+1 forecastlstm · gru · arima · ann
132.5133.0133.5134.0134.5FEBMARAPRMAYJUNJULAUG
actualmodel
EX.04Xuno · Kathmandu

Forecasting the Rupee

Remittance pricing lives and dies on tomorrow's rate

LSTM, GRU, ARIMA and ANN models in PyTorch forecasting USD→NPR, plus automated data pipelines on AWS that cut manual work by 30%. The model line hugs the actual line — that's the product.

98%
directional accuracy
+30%
pipeline efficiency
t+1
day-ahead forecasts
PyTorchLSTM/GRUARIMAAWS
sim // Five Apps, Five Weeks2024
deploy@headstarter — jul–aug 20241 prod deploy / week
WK 01
Inventory AI
Next.js · Firebase
○ queued
WK 02
Support Bot
OpenAI · Pinecone
○ queued
WK 03
Flashcard SaaS
Llama 3.1 · Groq · Stripe
○ queued
WK 04
ProfRank RAG
RAG · Embeddings
○ queued
WK 05
AI Capstone
Next.js · AWS
○ queued
users onboarded: 0+0/5 live
EX.05Headstarter AI · SF Bay Area

Five Apps, Five Weeks

Most side projects die at 80%. Mine had a weekly ship date

A production deploy every Friday for five straight weeks — a support chatbot with RAG over Pinecone that cut resolution time 50%, a flashcard SaaS with Stripe billing, and three more. Each scaled past 200 active users.

1,000+
combined users
–50%
support resolution time
5/5
weeks shipped on time
Next.jsOpenAIPineconeGroqStripe
sim // Bhanu: Saving Stories2024 —
uunchai oral archive · rec 047devanagari · nepali

Once, in the lap of the Himalayas, there was a small village — and a grandmother whose stories echoed through it…

10 students · one rule: don’t let the stories die
EX.06Uunchai (उँचाई) · Co-founder

Bhanu: Saving Stories

Nepali folk stories die with the grandparents who tell them

An NLP pipeline that archives and translates regional Nepali folk stories before they disappear — built by ten high-school students I recruit, train and lead. Technology pointed at memory.

10
students led
30+
workshop participants
stories worth saving
NLPPythonDevanagari corpus
01

ship weekly

momentum is a feature

02

measure what matters

users, not stars

03

bring people with you

teams outlast code

02range

Writes the PyTorch.
Reads the DCF.

Most engineers can't sit on the other side of the table. I trained for both — drag the divider.

fx_forecaster.py — xuno

import torch.nn as nn

class FXForecaster(nn.Module):
    def __init__(self, n_features=6, hidden=128):
        super().__init__()
        self.lstm = nn.LSTM(n_features, hidden,
                            num_layers=2, dropout=0.2,
                            batch_first=True)
        self.head = nn.Linear(hidden, 1)

    def forward(self, x):  # x: [B, 90, 6]
        out, _ = self.lstm(x)
        return self.head(out[:, -1])  # t+1 USD→NPR

# test: 98% directional accuracy · MAE 0.31 NPR

illustrative dcf — wharton case work

PV of FCF (’25–’29)412
PV of terminal value1,238
Enterprise value1,650
Less: net debt(210)
Equity value1,440

wacc 9.8% · g 2.5% · npr mm

engineerinvestor-literate
Warburg Pincus Fellow ’25 — 1 of 34 nationwide Wharton — 60+ hrs of DCF · LBO · comps · M&A AWS All Builders Welcome Grant ’25 Peer-reviewed post-quantum cryptography co-author

03leadership

Code scales.
People compound.

Every important thing I've shipped, I shipped with a team I built. Founding things is a habit, not a line on the résumé.

0+
students engaged
Moliss Tech Community — Founder

10 bootcamps, 7 workshops, 5 hackathons across 13 schools. Built from zero.

0
students led
Uunchai — Co-founder

Recruited and trained the team building Bhanu, the folk-story archive.

0K
people reached
Covid Resources Nepal — Research Lead

Led research & campaigns that moved $20K to 900+ families mid-crisis.

0
mentees shipped HamroGPT
Incubate Nepal — Peer Mentor

Five students, one real product, 5/5 satisfaction from real users.

350+ people taught, mentored, or led — before turning 22.

04also shipped

The rest of the ledger.

Public experiments — some scrappy, all real, all on GitHub.

+ olympiad honorable mention · + “vehicle for your route”, 200+ beta users · full ledger on github ↗

05ship log

Six years of receipts.

2021

  • Nepal’s second COVID wave. We ship a real-time hospital-bed & oxygen tracker — $20K raised, 900+ families supported, 84K reached.
  • Quantum computing research begins at OneQuantum.

2022

  • Found Moliss Tech Community — 22 events, 275+ students, 13 schools.
  • Co-author a peer-reviewed paper on post-quantum cryptography.
  • Equity Bridge internship: scholarship-matching dashboards, +50% successful matches.

2023

  • Kathmandu → Lancaster, PA. Computer Science @ Franklin & Marshall (’27).
  • HamroGPT ships at Incubate Nepal — +40% engagement, 5/5 satisfaction.

2024

  • ML engineering @ Xuno — 98% USD→NPR forecast accuracy in PyTorch.
  • Headstarter fellowship — 5 AI apps in 5 weeks, 200+ users each.
  • Co-found Uunchai; “Bhanu” begins preserving Nepali folk stories.

2025

  • Warburg Pincus Fellow @ Wharton — 1 of 34 nationwide. 60+ hours of DCF, LBO, M&A.
  • AWS All Builders Welcome Grant recipient.

2026

● now
  • 1Day begins — one person, one spotlight, every 24 hours. Waitlist is live.
  • SecAPI ships. GPA 3.75. Graduating May 2027.
  • Open to fall ’26 co-ops and new-grad ’27 SWE / ML roles. You’re early.

06spotlight

Code compiles.
It doesn't dance.

Music, fitness, markets — the parts of me that don't fit in a repo live on Instagram. The stage is dark. You brought a torch.

stage // @withbinayakafter hours

open — fall ’26 co-ops · new-grad ’27 · swe / ml

Building something
ambitious?

I'd love to talk. I reply fast — shipping weekly trains you to.

email me ↗