Computer Science (Cyber Security) undergraduate — building at the intersection of
machine learning, full-stack engineering, and security.

• B.E. in Computer Science & Engineering (Cyber Security) at MS Ramaiah Institute of Technology — CGPA 9.52 / 10.
• Project Intern at Samsung R&D Institute India (PRISM) — building and compressing a multimodal vision-language model for on-device emotional and contextual captioning.
• Co-author of a Springer Nature book chapter on Quantum Computing for IoT.
• I like taking hard problems end to end — ML systems, production web apps, and security tooling — with tests, CI, and measurable results.
• Programmer on LeetCode (rating 1821, top ~7%).

Reconciles a merchant's order ledger against a payment gateway's report and the bank statement — three systems recording the same money that never agree. A tiered deterministic matcher resolves what it can; a bounded agent investigates the rest behind a verification gate that requires every claimed match to resolve to exactly one settlement, tie exactly on amount, and fall inside the payout window. Ambiguity that can't be resolved escalates rather than guessing, because a wrong match in financial data is invisible.

• 90.8% resolved and 100% classification accuracy across 14 classes, graded against a ground-truth key the engine never reads. Across 12 independent batches: 92.7% ± 0.4% resolved, accuracy unchanged.
• 111 tests verified by mutation. CI asserts the exact accuracy figure, so a regression fails the build instead of quietly editing a number in the README. A separate job proves the engine still reconciles with no language model configured at all.
• Contested matches solved jointly with a hand-implemented Hungarian algorithm, verified against brute force on 300 random matrices.
• Scope stated honestly: the data is synthetic and generated by the same author, so the numbers show the engine catches the defect classes it was designed around — not performance on a real merchant's books.
• Python · FastAPI · pytest · GitHub Actions
• Live report · Live app · Five-minute walkthrough

An agent that finds, proves, and patches code vulnerabilities. An LLM surfaces candidate findings; a validator generates a proof-of-concept exploit and runs it in a network-disabled Docker sandbox, discarding anything it cannot prove. The secure-fix patcher sits behind a human-approval gate.

• A self-correction loop took recall from 60% to 100% at 100% precision across five vulnerability classes on a reproducible eval harness.
• Provider-agnostic model layer via LiteLLM (Ollama / Claude / OpenAI).
• Python · LiteLLM · Docker · Python AST · difflib · pytest · GitHub Actions

Full-stack detection platform for DGA, tunneling, exfiltration, and C2 beaconing. A 22-feature Random Forest and Isolation Forest ensemble feeds a SOAR auto-block/sinkhole layer mapped to MITRE ATT&CK, with SHAP-explained reports.

• Threshold calibration against a false-positive budget cut false positives from 407 to 21 at unchanged recall.
• Benchmarked on 12,000 domains across 4 DGA families. The Zeek exfiltration sensor is vendored third-party (BSD-3).
• Python · FastAPI · React · scikit-learn · Docker · Chrome MV3 · pytest · GitHub Actions
• Live demo: dns-sentinel.vercel.app

Constrained model predictive control for a production choke on a naturally flowing oil well — the controller holds setpoints under process constraints instead of relying on manual operator adjustment.

• Interactive simulation you can drive yourself: autonomous-choke-controller-v1.streamlit.app
• Python · MPC · Streamlit

Web app that turns a topic into a structured course outline, built with a production workflow rather than a demo one — typed end to end, tested, containerized, and running on CI.

• React · Next.js · TypeScript · Jest · Supertest · Docker · GitHub Actions

Browser-based 3D cube you can scramble and solve, using Herbert Kociemba's two-phase algorithm via cubejs. Getting the CommonJS solver into a modern ESM build meant writing a small require shim and bridging it at runtime.

• TypeScript · Vite · Three.js

• EmoCapNet (Samsung PRISM) — multimodal vision-language model for emotion-aware captioning, compressed for on-device use. Currently writing it up for ICASSP.
• Springer Nature book chapter — Quantum Computing for IoT.

Languages — C++, Python, TypeScript, JavaScript, SQL
ML — PyTorch, scikit-learn, Hugging Face Transformers, SHAP
Web — React, Next.js, FastAPI, Node.js
Infra & tooling — Docker, GitHub Actions, pytest, Jest, Linux

Foundations — Data Structures & Algorithms (C++), Operating Systems, DBMS, Computer Networks, OOP

Open to software engineering internships. The fastest way to reach me is email.