I build AI for networks. An Information Technology student who wants to take good systems into public service.
I study Information Technology at Maharaja Agrasen Institute of Technology, Delhi, and ranked first in my department for two consecutive semesters.
My internships have covered network engineering, IoT, investment banking and education. They taught me that good technology is only half the job. People, budgets and trust decide whether it gets used.
Outside class I manage a personal equity portfolio, read investing classics, and teach children who have little access to good schooling. Compounding is the thread through all of it: how I invest, how I train, and how I plan to work.
Drag the sliders. The dashed line is what you put in; the curve is what time does with it. My own portfolio has returned about 28% a year over three years, but past returns are not a forecast, so this starts at a sober 12%.
The same arithmetic works on habits. A year of tiny improvements is not a little better; it is a different person.
Designed a real-time monitoring tool for POP and node health in Python.
It combines traffic forecasting, anomaly detection by voting across Isolation Forest, DBSCAN and Z-score, and Q-learning route recommendations, all served through REST APIs. The lab below is a small, playable version of the voting idea.
Built smart-lighting analytics dashboards on ESP32 and 5G modules.
Worked with 100K+ IoT records and trained an ML fault-prediction model at 95% accuracy that helped cut maintenance cost by 35%.
Supported live deals and pitches.
Did financial modeling, valuation, comparable-company analysis and due diligence, and helped put together pitch books.
Started by writing content, then moved into the classroom.
Taught English, math and computer skills to underprivileged children at a teaching center. This is where my goals changed.
Analysed behavioural and transaction data.
Looked for anomalies in the data and built risk dashboards from what turned up.
Live traffic from an imaginary POP. Inject a fault and watch three detectors vote. Two or more votes raise an alarm. You can also click the chart to drop a spike exactly where you want it.
> listening…
An in-browser toy of the voting idea from my Tata Communications tool. The real one used Isolation Forest, DBSCAN and Z-score; here two lightweight stand-ins sit alongside the Z-score so it runs at 60 fps.
Reads contracts, summarizes them, translates them and scores risk from 0 to 100.
Buy, hold and sell signals with stop-loss, drawdown limits and Sharpe analytics, paper-traded on live market data.
Zone-wise health scores and repair-time analytics for a smart-lighting network.
Automates tweet scheduling, engagement analytics and hashtag optimization through APIs.
Everything I reach for, wired up by how I actually use it together. Grab a node and throw it around; filter to see one cluster light up.
Run academic, cultural and technical events, from sponsors and logistics to problems on the day.
Organised case competitions, guest lectures and consulting workshops.
Developed a business model and prototype for an AI and SaaS product.
Information Technology at MAIT, building AI systems for networks and markets along the way.
To learn the half of the job that technology alone does not solve: people, budgets and trust.
Working on education and women's welfare programs, where good systems reach the most people.
One that takes what works in government to places it doesn't reach.
Let's talk.
Open to conversations about technology, finance and education. Email is the fastest way to reach me.