B.Sc Mathematics @ University of Delhi Available for Summer 2027 internships
Hey, I'm Shivam — a Mathematics undergraduate at Shivaji College, University of Delhi, passionate about Machine Learning, Backend Engineering, and Quantitative Finance.
I enjoy building systems where mathematics, data, and software come together — whether it's developing scalable APIs, engineering backend infrastructure, training machine learning models, or exploring quantitative trading concepts. Most of my time is spent learning, building, and shipping projects that challenge me to think deeper and engineer better solutions.
Currently, I'm focused on Machine Learning, FastAPI, System Design, SQL, DSA, and quantitative research fundamentals while exploring how intelligent systems and data-driven decision-making can be applied in real-world environments.
Always building. Always learning. Always optimizing.
Led a 3-person team building a NASA TESS exoplanet detection pipeline: BLS transit detection + Random Forest classification, 97% accuracy, FastAPI backend, Next.js frontend. Submitted with full deck, technical report, and README.
Production-grade URL shortener with Redis cache-aside redirects, atomic click counting, sliding-window rate limiting, and real-time analytics — fully async.
End-to-end A/B testing platform on an embedded DuckDB warehouse — Delta Method variance correction for ratio metrics, Sample Ratio Mismatch guardrails, and a live Streamlit dashboard for concurrent conversion + revenue experiments.
Modeled the core tension in market making: an adverse-selection baseline plus a full Avellaneda–Stoikov (2008) implementation, run through 10k–15k path Monte Carlo simulations with P&L, inventory, and sensitivity analysis.
Exoplanet detection pipeline on real NASA TESS data — BLS transit detection feeding a Random Forest classifier, with a FastAPI backend and Next.js dashboard. Validated on WASP-126b, recovering its orbital period to within 0.013% of the published value.
97% classification accuracy · ISRO × Hack2Skill HackathonModels the core tension in market making: an adverse-selection baseline plus a full Avellaneda–Stoikov (2008) implementation, run through 10k–15k path Monte Carlo simulations with P&L, inventory, and sensitivity analysis.
Sharpe ≈ 0.45 · ~51% profitable pathsEnd-to-end A/B testing platform on an embedded DuckDB warehouse — Delta Method variance correction for ratio metrics, Sample Ratio Mismatch guardrails, and a live Streamlit dashboard for concurrent conversion + revenue experiments.
Production-grade URL shortener with Redis cache-aside redirects, atomic click counting, sliding-window rate limiting, and real-time analytics — fully async and containerized.
Regression model predicting individual medical insurance charges from demographic and lifestyle features.
I'm quick to reply and always happy to talk about a project, a role, or just math.