Final-year Computer Science student at BRAC University, Dhaka. I work on reliable machine learning and quantum computing: careful benchmarks, calibrated uncertainty, and honest comparisons between classical and quantum models.

• 🔬 Current focus: variational quantum circuits, conformal prediction, and continual learning.
• 🎯 Aspirations: graduate study (M.S. / Ph.D.) in AI/ML and quantum computing.

Each of these began as a result and turned into a finding about the method behind it. The controls were built so that they could fail, and where they did, that is what the repository reports.

• crop-fusion-bench — image and greenhouse-sensor fusion for crop-disease diagnosis, with a parameter-matched 8-qubit VQC head against its classical twin. The headline is an audit of the benchmark itself: 98% of test samples share a sensor recording with training, and a lookup table on that recording reaches 94.0% test accuracy with no learning at all, against the trained sensor model's 94.4%.

• prototype-aggregation-cl — which prototype rules keep a continual learner from collapsing, over 120 controlled cells. Projecting into 10,000 dimensions does not prevent collapse, and bounded aggregation alone only partly does; hyperdimensional computing matches an exact-median learner while keeping one vector per class instead of every sample it has seen.

• hcc-kinase-conformal-screen — conformal applicability domains for virtual screening of natural products against three liver-cancer kinases. The domain filter turns out to depend on a tolerance parameter that published work rarely states: at the value used here it switches itself off on two of three targets, admitting the whole library as "in domain".

• Quantum Physics Lab — 84 interactive lessons in five tracks, in the browser, in vanilla JavaScript with zero build dependencies: wave-particle duality, Bell tests, Bloch sphere visualisations and surface-code error correction, with experiments run on IBM's ibm_kingston processor. Live site

• Compiler Design — a C compiler frontend: lexical analysis (Flex), LALR parsing (Bison), scope-nested symbol tables, type checking, and three-address code generation.

• Personalizer — offline-first PWA workspace for academic tracking and credential management, with a stale-while-revalidate service worker and Firestore persistence.

• SiteForge — construction resource planning: inventory tracking, procurement workflows, BOQ generation and field workforce management, on Node.js, Express and MySQL.

• Vector Velocity — a 3D arcade racing engine written from scratch in Python and PyOpenGL, with procedural tracks, raycast collision physics and waypoint-driven AI opponents.

• Email: ishmam.rashid.bhuiyan@g.bracu.ac.bd
• GitHub: @thinkiamparanoid2