Christian Juresh

London, UK

I’m a software engineer on an integration team, shipping work end to end — backend, platform and frontend.

Clivia in bloom
30 St Mary Axe, London
Kew Palm House glasshouse
Red pomegranate blossom
Dome of St Paul’s Cathedral
Hanging catkins, Kew Gardens
Riverside towers through a telescope
White blossom
Red Thames Clipper on the river
Spiral aloe succulent
Tower of London turret with a Union flag
White umbel flowers
Domed building framed through a window
Pink anthurium flower
Curving railway tracks
Cherry blossom branch
White observation tower against the sky
Sunset over the rooftops
Cross-shaped arrow slit in a stone wall
Ornate turret with flowers
Treetop walkway, Kew Gardens
Port of London Authority building beside a crane
Shipping cranes at the docks
Bare tree against a blue sky
The Shard with a passing plane
The moon in a dark sky
Statue of an archer atop a tower
Contrails over rooftops at dusk
Glass towers of the City of London
Hazy countryside view
The Shard beyond rooftops, with a train
The moon above a tree at dusk
Canal and skyline
St George Wharf towers, Vauxhall
The London Eye across the Thames
Riverside skyline from the water
A single rose in low light
Kew Palace and lawn
citizenM hotel facade
Red flower in shade
White umbel flowers
White blossom
Red-brick lodge, Kew Gardens
Glasshouse across the gardens
Deckchair in the park
Kayakers on the river
Modern glass building
White office building
City office buildings
Countryside view
Countryside view
The London Eye from the water
Glass roof, London

Work Experience

Third Bridge Group LimitedThird Bridge GroupThird Bridge2024–Present

Royal College of RadiologistsRCR2022–2024

Education

BSc (Hons) Computer Science, QMULBSc Computer Science, QMULBSc Comp Sci, QMULComp Sci, QMUL1st · 82%2021–2024

Contact

Projects

Projects

Self-stackingphoto gallery

Twenty years of photographs had turned into 1,374,328 file paths across overlapping backups. This is the system that turned them into one immutable content-addressed vault and a website that browses it. When the shipped result disagreed with what I actually wanted, the objective moved and the reasoning that produced the old one stayed on the page.

  1. Label-driven calibration

    96.5% precision, 85.6% recall — measured, not chosen

  2. Screen-then-verify matcher

    3,632,211 candidate pairs, screened to 566,522

  3. Ledger-backed delete gate

    435.6 GB re-hashed and reconciled both ways

  4. Triage rule engine

    1,374,328 paths re-costed in 220 ms