pratham@thapar:~$ whoami
3rd-year CSE. Trains models, ships them, occasionally organizes festivals for 5,000 people.
pratham@thapar:~$ cat account_history.log
[May 2025] GitHub account created. first-repo, in C. Humble beginnings.
[Since] 26 repos · a package on PyPI · a booking site a real business runs on
· Amazon ML Summer School (80,000 applied, ~3,000 in — still think about that email)
pratham@thapar:~$ grep "origin story" .
Not the "started coding at 8" type. The type that shows up every day.
pratham@thapar:~$ ./looking_for
Summer 2026 AI/ML internship — somewhere I can ship real things next to people better than me.portfolio · linkedin · kaggle · pgarg7_be23@thapar.edu
Every one started as a real problem — mine or somebody's. Each comes with a schematic and the honest part — that's where the actual engineering lives.
Yogic — can a model tell you're stressed before you can?
A Mixture-of-Experts network that reads stress off four biosignals. Each expert learns one sensor's patterns; a gating network fuses them into a continuous stress score. Python · TensorFlow · time-series
the honest part ↓
Four sensors, four different sampling realities. Getting the streams aligned and the experts to actually specialize took longer than the model itself. Next time: attention-based encoders per sensor and a proper ablation of the gating network.
The Pickleball Arena — a real booking site, for a real business, in my hometown
Court bookings for a pickleball facility in Bathinda, Punjab. Not a portfolio demo — actual players reserve actual courts. React · Vercel · live
the honest part ↓
The users aren't tech people — they're players booking from their phone between games. Every extra tap costs a booking. Next: the coaching module. The site says "launching soon," so I should probably launch it.
Happy Hoping — born on a trek. true story
Trekked to Jibhi, couldn't find one decent café recommendation online, came home and built the fix. A food-discovery engine that learns your taste instead of pushing sponsored lists. JavaScript · recommendations · live
the honest part ↓
Small-town India barely exists in food-data APIs. That gap is literally why the project exists. Next: a real preference model instead of heuristics, and more towns.
CogniAI — teaching cameras to sort a city's garbage
Real-time computer-vision waste classification for smart-city recycling, built with Sanyam Wadhwa. Python · CNNs
the honest part ↓
"Accurate" and "fast enough for real-time" pull in opposite directions; most of the work was in the middle. Next: edge deployment and a feedback loop from misclassifications.
TOPSIS on PyPI — my first published package. roll number included.
pip install Topsis-Pratham-102303052 # yes, reallyA working TOPSIS implementation on the real Python Package Index — versioning, docs and all. Python · MCDM · published
the honest part ↓
The method took an afternoon. Packaging, metadata and versioning took the weekend.
More experiments: transformer (grid-cell targeting for distribution transformers — CLI + folium maps) · numerai · campuspo (Next.js, in progress) · neetcode-submissions (the daily grind, in C++)
| daily drivers | Python · Pandas/NumPy · scikit-learn · SQL · C++ (DSA) · Git |
| building with | TensorFlow · PyTorch · FastAPI · Docker · Postgres |
| actively learning | LLM APIs & RAG · vector DBs (FAISS/Chroma) · Next.js · MLOps |
Tiers, not percentage bars — nobody knows what 80% of Python means.
pratham@thapar:~$ cat now.txt
→ studying : 3rd year CSE, Thapar
→ building : campuspo · this README, apparently
→ grinding : 30 Nites of Code · post-50-day-badge LeetCode streak
→ training : run · lift · planning the next trek
→ looking : Summer 2026 AI/ML internshippratham@thapar:~$ ./stats --since "May 2025" --no-excuses
Every green square since May 2025 — one year of account, no vintage commits to pad it out.
Run, lift, trek, repeat — progressive overload works on muscles and on ML skills. Helped organize fests for 5,000+ attendees (a full year on VIRSA's exec committee). One season I was even an actor and model. Long story — ask me.
timeline, if you're curious ↓
- Aug 2023 — started B.Tech CSE at Thapar; joined VIRSA exec committee within weeks
- Nov 2024 — Saturnalia fest: core member and actor/model ★
- May 2025 — created this GitHub account
- Aug 2025 — selected for Amazon ML Summer School '25
- Jan 2026 — published TOPSIS to PyPI
- Jan–Feb 2026 — built CogniAI and Yogic
- Now — you, hopefully →
Account created May 2025. Everything above happened in one year. Imagine year two.
Powered by chai and Ctrl+R. Punjab raised — Bathinda → Patiala.