Software Engineer

Joshua Dardashti

I build data-intensive systems in Python and C++.

Trading research infrastructure, orbital mechanics engines, event-streaming pipelines, and database performance tooling. B.S. Technology and Information Management, UC Santa Cruz, June 2026. Open to backend, data, and infrastructure engineering roles.

View GitHub Resume (PDF) Get in touch

Selected work

Space-Domain Awareness Collision Predictor

Public

Satellite conjunction analysis engine. A single collision-probability estimate is easy to get wrong and hard to notice when it is, so this engine computes probability three independent ways, Foster, Chan, and Monte Carlo, and cross-validates them before a result is trusted. SGP4 propagation feeds a two-phase screen, results ship as CCSDS-format conjunction data messages, and a live 3D catalog globe is served over FastAPI.

PythonSGP4FastAPICesiumJS
View source
  • +Over 200 automated tests
  • +Three probability methods, cross-validated
  • +CCSDS conjunction data message generation
  • +Live 3D satellite catalog globe
  • +Maneuver planning and screening thresholds

Crossbook

Public

Multi-venue crypto order book reconstruction in C++20. Order book code usually assumes it is correct; Crossbook proves it, verifying the rebuilt Kraken and Binance L2 books against exchange checksums on every update from the live streams.

  • +Checksum-verified on every book update
  • +Header-only feed handlers, two venues
C++20Market dataLow latency
View source

Alpha Research Platform

Source private

Systematic trading research platform. Multi-source ingestion lands futures, macro, and derivatives data in a 7.6 GB Parquet lake partitioned by dataset, schema, and symbol, with per-fetch manifests recording row counts, cost, and provenance. DuckDB SQL runs directly over the lake; a NATS JetStream event backbone provides durable, idempotent delivery with end-to-end trace context.

  • +7.6 GB partitioned Parquet lake with provenance manifests
  • +Vectorized backtests cut from minutes to seconds
PythonDuckDBPolarsArrowParquetNATS

SQL Server Plan Lab

Public

Five SQL Server performance pathologies, each reproduced with one command against a seeded multi-million-row database. Every published number is regenerated by CI: if a pathology stops reproducing, the build fails.

View source

Alien Escape

Public

The 1981 Galaga arcade rules rebuilt as pure, unit-tested ES modules behind a thin Phaser 3 rendering layer, so every behavior is testable headlessly. An unofficial fan tribute.

View source

Skills

Languages

  • Python
  • C++20
  • SQL / T-SQL
  • JavaScript

Data and Infrastructure

  • DuckDB, Polars, Apache Arrow
  • Parquet data lakes
  • NATS JetStream
  • FastAPI
  • SQL Server, SQLite
  • NumPy, Pandas

Practices

  • Test-driven development, pytest
  • CI with GitHub Actions
  • Docker-based reproducibility
  • Walk-forward validation
  • Git

Education

University of California, Santa Cruz
B.S. Technology and Information Management, June 2026
GPA 3.9

Let's talk.

Open to backend, data, and infrastructure engineering roles. The fastest way to reach me is email; I typically reply within a day.

jsdardashti@gmail.com LinkedIn