Hi, I'm Arnab, a Cornell CS & Mathematics student and Compiler Engineering Intern at Intel. I build reliable software systems with an emphasis on formal methods, compilers, and programming languages, then carry those ideas into practical tools that people can actually ship and use. I care about correctness, mathematical elegance, sharp developer workflows, and systems that stay maintainable as they grow.
Current Work
Intel
I work on the PTA team, hunting for performance headroom in the icx compiler across C/C++/Fortran workloads — things like SEAL (homomorphic encryption), LLVM and CPython builds, zstd, discrete-event network simulators, and an EDA chip-design tool. I use perf and VTune to chase a regression down to a minimal repro, then write and patch optimization and simplification passes, debug vectorization failures, and tune inlining/unrolling to actually recover the performance. I found and fixed a hardware-level TLB-miss regression, and wrote a new vectorization pass plus hand-vectorized NTT/inverse-NTT kernels that pushed throughput on the affected benchmarks up by 10%.
Credentials
Cornell University
Aug 2024 – May 2028
GPA: 3.81/4.0
Emphasis on programming languages, compilers, runtime verification, formal verification, and applied logic. Also majored in mathematics, focused on different logical systems and algebras.
Coursework & Activities
Professional
Hypercubic
Focused on specialized verification tooling, domain-specific languages, and compiler infrastructure.
HRL Laboratories
Built and operated quantum-computing research infrastructure using Python, TypeScript, Vue.js, Docker, GitLab, Artifactory for researchers in secure air-gapped environments, supporting spin-qubit adjacent workflows and improving end-to-end deployment, throughput, and usage organization-wide.
Cornell Course Management System X
Lead for CMSX, Cornell's in-house course management system serving thousands of students across hundreds of courses each semester. Modernizing tightly coupled XML and legacy JSP to modern Java REST APIs and type-safe React frontends, and built the platform's observability stack from scratch — Grafana dashboards, Prometheus with a custom MySQL exporter, Loki/Alloy log aggregation, and severity-tiered Slack alerting for incident detection.
Cornell University
Led advanced discussions on JVM internals—bytecode execution, JIT/tiered compilation, and hotspot behavior—and taught formal grammars and AST-based language implementation; mentored students building a programming language runtime.
Selected Work
All projects →Quon
A Rust/MLIR optimizing compiler toolkit for quantum programs: a linearly-typed source language with symbolic depth bounds and inferred Clifford classification, lowering through custom MLIR dialects to OpenQASM 3 (SABRE routing, native-gate decomposition) or to verified neutral-atom schedules with a QEC workload IR, ZX-calculus and Clifford+T rewriting, and Stim/Sinter-validated resource reports. 100K+ lines of Rust, 500+ commits, with its own formatter, linter, LSP, and release toolchain.
Owl
Runtime verification for Gleam using sHML-based monitors compiled from a temporal property DSL, with low-overhead BEAM bytecode tracing for distributed actor systems. Started as a capstone for CS 6156; now growing toward causal, evidence-backed verdicts over cross-system action sequences rather than a single BEAM app.
Fern
A language targeting LLVM IR with Hindley-Milner + Algorithm W type inference, refinement types, trait constraints, row polymorphism, and exhaustive pattern matching.
Bo Nix
Full optimizing compiler for Eta/Rho → x86-64 in Java: lexer through register allocation, ABI-compliant codegen, and language extensions including records and null handling. Capstone for CS 4120.
Guardrails: Atomic
Built a Next.js 15 + TypeScript platform where engineers define distributed-system requirements in a YAML DSL and run a Z3-backed CEGIS loop with GPT-4-guided synthesis/repair until invariants are formally discharged.
Technical Writing
All writing →