Software Engineer • Machine Learning & NLP • Research-Oriented

LinkedIn • Email • Codeforces

Most people ask ML systems:

“What’s the prediction?”

I usually ask:

“Should we even trust this prediction?”

I’m an Integrated M.Tech (CSE – Computational & Data Science) student at VIT Bhopal,
interested in building machine learning systems that behave responsibly, explain their limits,
and don’t pretend to know more than they actually do.

I enjoy problems where correctness, diagnostics, and reasoning matter more than flashy accuracy numbers.

• Teaching models when to speak and when to stay silent
• Designing end-to-end ML pipelines with clean evaluation and guardrails
• Building NLP systems for large, real-world legal text corpora
• Exploring cryptography and ternary logic where assumptions really matter
• Turning messy ideas into readable, reproducible systems

Languages
Python • Java • C++ • SQL

Machine Learning & NLP
Scikit-learn • TF-IDF • SVM • Feature Engineering • Model Evaluation • Model Diagnostics

Tools
Git/GitHub • Streamlit • Selenium • Tkinter • AWS

Foundations
Data Structures & Algorithms (300+ problems) • DBMS • OOPS

A market analysis system that refuses to act confident when it shouldn’t.

What it does:

• Produces 5-day probabilistic outlooks instead of single guesses
• Detects market conditions (calm vs unstable) before interpreting signals
• Explicitly rejects assets and predictions that are statistically unreliable
• Stress-tests model behavior during extreme events (e.g. COVID-19 crash)
• Explains results in plain language, not just metrics

What it doesn’t do:

• It doesn’t promise profits
• It doesn’t hide uncertainty
• It doesn’t pretend every signal is useful

📌 Built as a learning and diagnostic system, not a trading bot.

🔗 https://github.com/Ayush-1271/MarketLens

NLP system for classifying 5,000+ Indian Supreme Court judgments

• 93% accuracy using TF-IDF + SVM
• Reduced manual classification effort by ~70%
• Focused on scalability, interpretability, and evaluation rigor

Attendance system using face recognition + GPS

• Prevents proxy attendance via multi-factor validation
• Designed with real-world deployment constraints

Automated web scraping system with GUI

• Converts web novels into structured PDFs
• Multi-threaded, fault-tolerant, and user-focused

Cryptographic primitive built using ternary logic

• Evaluated using NPCR, UACI, entropy analysis
• Paper currently under academic review

• Correctness beats shortcuts
• Diagnostics beat blind accuracy
• Clear reasoning beats clever tricks
• If a model is unsure, it should say so

I’m especially interested in ML systems, NLP, and research-oriented engineering roles
where judgment matters as much as code.