ISP RAS — SharpChecker development
- Work on symbolic computation in SharpChecker, a C#/.NET static-analysis tool.
I build language infrastructure that separates semantic decisions from execution. I currently work on symbolic computation in SharpChecker at ISP RAS; UniversalToolchain fixes typed backend routes in a checked plan, while PlanFuzz probes configuration failures.
2022: Compilers / VMs · 2023: CIL + native x86-64 · 2024–2025: AST / IR / backends → toolchain · 2026 →: LLVM + program analysis
Resolve dependencies, conflicts, and backend routes into an immutable LanguagePlan. An independent verifier checks the plan; runtime does not repeat semantic selection.
Current manifest: 1,324/1,324 tests; dedicated checks cover interpreter/CIL parity and configuration integrity before execution.
Experimental testing of programs and configuration plans: isolated replay, control/fault cases, and counterexample reduction that preserves failure fingerprints.
Frozen fault/control corpora and replayable artifacts make planning defects inspectable without claiming formal proof.
An LLVM 22 pass localizes supported induction-like globals after interprocedural effect analysis, affine evolution, and a separate applicability check; APInt models fixed-width wraparound.
29 positive/negative regression cases cover LLVM Verifier checks, structural expectations, idempotence, and observable before/after behavior.
Compiler Infrastructure / Language Tooling Engineer
I build language infrastructure that separates semantic decisions from execution. I currently work on symbolic computation in SharpChecker at ISP RAS; UniversalToolchain fixes typed backend routes in a checked plan, while PlanFuzz probes configuration failures.
Compiler project lineage
01 · Experience
C#/.NET · static analysis · symbolic computation
LLVM 22 · C++23 · compiler engineering
02 · Projects
Resolve dependencies, conflicts, and backend routes into an immutable LanguagePlan. An independent verifier checks the plan; runtime does not repeat semantic selection.
Current manifest: 1,324/1,324 tests; dedicated checks cover interpreter/CIL parity and configuration integrity before execution.
An LLVM 22 pass localizes supported induction-like globals after interprocedural effect analysis, affine evolution, and a separate applicability check; APInt models fixed-width wraparound.
29 positive/negative regression cases cover LLVM Verifier checks, structural expectations, idempotence, and observable before/after behavior.
03 · Skills
dependency/conflict resolution, component/provider selection, deterministic pass ordering, typed artifact routing, pre-execution checks
C++23, LLVM 22/IR, CFG/SSA, loop and interprocedural analysis, transformation applicability
.NET, CIL/MSIL, native x86-64, SysV ABI, stack frames/calling conventions, register allocation, interpreter/runtime composition
regression suites, LLVM Verifier, interpreter/backend parity, differential execution
04 · Recognition
Organizer correspondence: only 5.0/5.0 score and only solution passing all 104/104 tests among 49 submissions.
Delivered the UniversalToolchain architecture talk at LangDev on 9 October 2026.
Selected as one of the winners with UniversalToolchain / Wist2.
MEPhI Junior: Information Technologies section winner in 2025 and first-degree diploma in 2026; HSE Vysshaya Proba Industrial Programming diploma holder; Baltic Science and Engineering Competition: first-degree diploma and the “Perfection as Hope” prize.
05 · Education
Moscow