Stanislav Kolesnikov

Open to python & full-stack roles

Engineering new languages, compiler pipelines, low-level systems and native environments.

// HIGH PERFORMANCE PIPELINES // COMPILER DEV

01 // ARC

10 ★

Arc is a lightweight programming language written in C, featuring a custom lexer, parser, and interpreter. It supports variables, functions, control flow, and a minimal standard library.

Case Sensitivity Matrix

Keywords are Case-Insensitive: Control frameworks like VAR, FN, IF, WHILE, TRY, and RETURN accept any matching casing standard dynamically.

Identifiers are Case-Sensitive(myVariable and myvariable are separate instances).

Native Literal Evaluation

  • Numbers: Maps both raw Integer values and IEEE floating-point designations.
  • Booleans: Truthiness is evaluated via isTruthy(Value v). Integers and floats evaluate false when equal to 0 or 0.0. null is also false. Objects are truthy except: empty strings and empty lists.
  • Strings: Immutable character sequences contained inside standard double quotes.

Object System & Explicit Contexts

Classes are built utilizing the CLASS keyword. Methods skip hidden structural context binding overlays - meaning instances have to get fed explicitly through the argument frame to process scope modifications.

CLASS Entity
    VAR field = "abc"
    FN func(instance, updatedValue) THEN
        instance.field = updatedValue
    END
END

VAR actor = Entity()
actor.func(actor, "new_state")

Low-Level VM Pipeline Architecture

  1. Lexical Stream: src/lexer.c parses source inputs into distinct structures.
  2. Recursive Descent Tree: src/parser.c produces a optimized ASTNode tree.
  3. Bytecode Emission: src/compiler.c targets memory arrays called Chunks while resolving variable scopes.
  4. Labels-as-Values Loop: src/vm.c runs instructions through an optimized stack framework utilizing computed gotos.
Memory Arena Allocation

Leverages src/memarena.c for persistent AST phases alongside custom recycling pools in src/mempool.c to reduce structural overhead.

Standard I/O Array

Natively provisions operations such as print(), len_of(), and math files like math.arc via explicit runtime inclusions.

C Language Compiler Design Bytecode VM Interpreter Check Out →
// NETWORKING ECOSYSTEMS // CORE ASGI

02 // NEBULA

9 ★

An asynchronous Python web routing engine built completely over native ASGI. Designed from the bottom up to serve direct data-streaming microservices without introducing the bloat and middleware overhead.

// PARALLEL EVENT LOOPS & ARCHITECTURE

Nebula sidesteps complex multithreading frameworks by executing over standardized state events. It provides micro-routers with flat lookups, achieving predictable throughput for real-time web applications.

from nebula import Nebula, run_prod

app = Nebula("127.0.0.1", 5000)

@app.get("/")
async def home():
    return {"status": "routing_active", "engine": "nebula_core"}

if __name__ == "__main__":
    run_prod(app, workers=4, access_log=False)
Python ASGI Framework Async Web Server Web Check Out →
// BARE METAL ASSEMBLY // LOW LEVEL RESCUE

03 // ARCOS

4 ★

A lightweight, real-mode 16-bit operating system written entirely in x86 Assembly. Features a multi-stage sector bootloader, native console shell interfaces, and low-level interrupt hooks communicating directly with hardware.

Interrupt 0x80 Register Interface

Programs communicate with the underlying kernel by preloading explicit identifiers inside the ah register block before issuing interrupt calls:

Syscall Tag AH Op Parameters Functional Objective
sys_print_char 0x00 al = character byte Writes an individual token directly to the screen view.
sys_print_string 0x01 si = pointer reference Processes null-terminated text streams to terminal screens.
sys_print_newline 0x02 None Moves cursor to new line.
sys_shutdown 0xA0 None Signals APM parameters to shut down system operations.

Integrating Custom Assembly Extensions

Adding custom utilities involves building targeted instruction sets inside the program registry before wiring them into the system layout checks.

; target logic written inside progs/custom_print.s
user_extension_entry:
    mov si, display_message_pointer
    mov bl, 0x0F  ; Light white text attributes
    call print_string_utility
    ret
display_message_pointer: db "Custom User Vector Triggered.", 0
NASM Assembly Operating System x86 Real Mode Low-Level Systems Check Out →
// INTUITIVE AI INTERFACES // FULL-STACK EDTECH

04 // EDUDUCK

6 ★

An AI-assisted study companion application that transforms handwritten logs and document streams into customizable training tools. Features automated quiz generators, summaries, and OCR for PDFs and images.

// APPARATUS & COMPONENT SPECS

Document Ingestion

Transforms inputs (including TXT, PDF, and image text extractions) into formatted reference frameworks via OCR preprocessing layers.

Intelligent Schedule Generators

Maps dynamically updating study outlines by analyzing target timelines, specific files, and personal learning preferences.

DuckAI Interaction Core

Maintains persistent message contextual tracking, giving users the power to converse with files and query complex technical configurations.

Auth & Localization

Google OAuth authentication with multi-language user interface support.

Python AI Application OCR Pipeline Full-Stack App Check Out →

// PRODUCTION RUNTIME PERFORMANCE TELEMETRY

HTTP BENCHMARK // wrk -t4 -c900 -d30 NEBULA CORE
Running 30s test @ http://127.0.0.1:5000
  4 threads and 900 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    20.72ms   35.79ms   1.16s    99.40%
    Req/Sec    12.18k     1.18k   16.08k    69.92%
  1455071 requests in 30.05s, 194.27MB read
Requests/sec:  48418.20
Transfer/sec:      6.46MB
ARC VS CPYTHON BENCHMARK COMPILER METRICS
single-threaded, warm cache, best of 5 runs

Fibonacci (recursive, n=30)     | Arc faster (~1.49x)
Fibonacci (iterative, n=1000)   | Arc faster (~2.38x)
Sum 1..1,000,000                | Arc faster (~2.47x)
String concat (x10,000)         | Python faster (~4.34x)
List build (x1,000)             | Python faster (~1.25x)
Nested loops (1000x1000)        | Arc faster (~1.96x)
Recursive countdown (5000)      | Arc faster (~2.14x)
Variable churn (x100,000)       | Arc faster (~2.19x)

Arc wins 6/8 benchmarks


// DAILY DRIVERS

OS: Arch Linux x86_64
Host: ASUS TUF Gaming F15
Kernel: Linux 7.0.12-arch1-1
Uptime: 3h 49m
Packages: 1081 (pacman)
Shell: zsh 5.9.1
Display: 1920x1080 @ 144Hz
WM: Hyprland (Wayland)
Terminal: kitty 0.47.1
CPU: i5-10300H (8 threads)
GPU: GTX 1650 + Intel UHD
Memory: 9.76 / 15.45 GiB
Swap: 1.15 / 7.73 GiB
Disk: 126.77 / 197.92 GiB
Battery: 100% (AC)
OS: NixOS 25.11 (Xantusia)
Host: ThinkPad L460
Kernel: Linux 7.0.9-zen1
Uptime: 2h 31m
Packages: 1230 (nix-system)
Shell: zsh 5.9
Display: 1920x1080 @ 60Hz
DE/WM: i3 (X11)
Terminal: kitty 0.44.0
CPU: i5-6200U (4 threads)
GPU: Intel HD Graphics 520
Memory: 1.60 / 7.63 GiB
Swap: 0 / 12.61 GiB
Disk: 17.04 / 209.35 GiB
Battery: 77% (AC)
// GET IN TOUCH

Have a project in mind or want to collaborate? Reach out.

© Stanislav Kolesnikov 2026