No. 4The Architect

Abdullah Ashmawei

AI solutions architect. Maps the design space before the code, across autonomous networks, geospatial intelligence and football game analysis.

Cultures
  • Arabic
  • English
  • Spanish
Roots
Egypt
Based in
Egypt
Portrait of Abdullah Ashmawei
A. ASHMAWEI
  • 4 Intelligence domainsAI architecture, autonomous networks, geospatial and football intelligence
  • 4 Years of football studyGame analysis, set pieces and football AI
  • 10 Professional journeysCountries in his career entries, from Japan to South Africa
  • 3 CulturesArabic, English and Spanish
  • AI solutions architecture
  • Design spaces
  • Evidence and decision
  • Autonomous networks
  • Geospatial intelligence
  • Football intelligence
  • Set pieces
  • Game analysis
  • Football AI
  • Closed-loop systems

Profile

Abdullah Ashmawei wears number 4, the architect’s shirt. He is the squad’s AI solutions architect: he decides what gets built, in which order and against which proof, before anyone writes a model.

Solution architecture

Every real problem has several honest architectures. He lays them side by side, strikes the ones the constraints cannot afford, and builds what is left.

Autonomous networks

His specialisation is AI solutions architecture for autonomous networks: systems that observe themselves, diagnose faults and act without waiting for a person.

Geospatial intelligence

Networks are maps. Coverage, demand and movement all live in coordinates, and so does every event on a football pitch.

Football intelligence

Four years of professional education in football game analysis, set pieces and football AI. A dead ball is where the two halves of his work meet: repeatable, measurable, the closest thing football has to a solvable problem.

Highlights

  1. The design space first

    Before a model is chosen, every viable route goes on one page, with what it needs and what breaks it.

  2. Evidence apart from decision

    What a frame says and what the match means are separate machines, each tested alone.

  3. Constraints as design inputs

    Privacy, hardware and reproducibility are requirements, drawn in from the first line.

  4. Decided on evidence

    The final call is made on data the design has never seen, by a rule written in advance.

  5. One architect, two domains

    The architect's habits and the analyst's questions in one person, so football problems arrive framed as whole systems.

The design space

Several honest routes. One decision.

  • Video onlySet aside: one laptop, a few days.
  • Hybrid fusionSet aside: one laptop, a few days.
  • Evidence + exact decodeFielded.
  • One learned classifierSet aside: no labels on the match.
  • Learned decodersSet aside: no labels on the match.
  1. FrameOne target: home, away or dead on every frame.
  2. CompareThree sources of evidence and seven decoders, on one grid.
  3. ConstrainOne laptop. No labels on the match. The same bytes on every run.
  4. DecideFootball evidence on every frame, one exact decode per half.

An architecture, layer by layer

The squad's decoder, from constraints to proof.

  1. ProofOpen DFL labels, the official scorer, both metrics
  2. SelectionPareto front of two objectives, frozen profiles
  3. DecisionExact Viterbi decode per half: home, away, dead
  4. EvidenceBall control, out of play, stillness, restarts, calibrated tags
  5. ObservationTracking and tags; a missing ball is no evidence
  6. ConstraintsOffline, one laptop CPU, deterministic, no labels on the match

Autonomous networks

People set the goals. The loop does the rest.

  1. ObserveEvery node reports.
  2. DiagnoseThe fault is located.
  3. ActTraffic takes another path.
  4. VerifyThe fix is confirmed.

Geospatial intelligence

Three layers of a map, then the pitch.

  1. CoverageWho reaches where.
  2. DemandWhere the load sits.
  3. MovementHow it flows.
  4. The pitchThe same questions, 105 by 68 m.

Football intelligence

Three ways he reads a match.

  • Game analysis

    Shape, lanes and the space a team leaves.

  • Set pieces

    Runs, screens and a delivery into the target zone.

  • Football AI

    A state on every frame, even when the ball is out of sight.

Body of work

Architecture

The design space of possession inference, drawn option by option, with the calls the squad made and the constraint behind each one.

16 studies in 5 families

Enter Architecture
Diagram: Observe, Evidence, Decode, Adjudicate, Prove.
  • Diagram: Tracking and tags, Alive or dead, Which team, Viterbi per half, H / A / D, Profile.

    On the pitch

    The Shrewdies decoder

    The architecture we fielded, end to end: football evidence on every frame, an exact Viterbi decode per half, and every parameter chosen away from the match.

  • The map

    The single source of truth

    Three witnesses watch one match: the video frames, the tactical tracking and the broadcast tracking. Each has gaps the others can fill. This is the architecture that reconciles them into one match state per instant, in which every value remembers which witness supplied it.

  • The engine

    Seven ways to decode a match

    The same per-frame scores, seven different stories. What each decoder assumes about football, what it needs, what it guarantees and how it fails.

  • The hard parts

    When the ball goes missing

    What should a design believe on a frame with no ball? Three treatments of the blank frame, and two schools of evidence: measure the ball when it is there, or infer possession from how the players move.

Professional journeys

  1. Egypt
  2. Sudan
  3. Yemen
  4. Saudi Arabia
  5. Qatar
  6. United Arab Emirates
  7. Türkiye
  8. South Africa
  9. Philippines
  10. Japan
Today
Dua, his current company
Journeys
From Japan and the Philippines to South Africa, Sudan and the Gulf

Profile facts: his public LinkedIn profile (current company and career locations) and the team's own records. The design space and the layers are those of the squad's Initial Challenge solution, documented in Architecture.