DANIEL
SHDEED.
Let’s talk
FULL STACK ENGINEER / INDEPENDENT RESEARCHERA CONTINUOUS WORK IN PROGRESS

Engineering
the unknown.

I build systems.
I study how they think.

From AI agents to the bytes inside a model. A collection of things I’ve built, experiments I’ve designed, and questions I’m still chasing.

Step inside the work
BUILT TO BE EXPLORED.
DESIGNED TO BE QUESTIONED.
SYSTEMS ENGINEERINGAPPLIED AIMODEL INTERNALSCONTROLLED EXPERIMENTS
01 / THE PROJECT COLLECTION

Ideas, made
tangible.

Step into the interfaces, look under the hood, and see where the evidence ends.

01 / REAL-TIME SYSTEMSBeta
Race desk · live-site replay capture

FastPitStop

Race intelligence. Without the spoilers.

A second-screen race companion where timing, strategy, and AI explanations follow the viewer’s clock.

SvelteKit / Python / TimescaleDB
Source ↗
02 / SPATIAL REASONINGPilot pending
Instance inspector · mock-provider run

Shadow Twins

Same shadows. Different structure.

A spatial benchmark with exact scoring. Inspect the original object, a model’s answer, and the certified optimum side by side.

Python / Three.js / Exact solver
03 / APPLIED AILocal implementation
Local development interface · seeded sign-in screen

OpsPilot

Every action needs an evidence trail.

An operations agent that investigates shipment exceptions, grounds its findings, and keeps a supervisor in control.

FastAPI / React / pgvector
Source ↗
04 / MODEL INTERNALSNear completion
Actual CLI output · intentionally unhealthy fixture

biopsy

Open the model. Understand the bytes.

A Rust CLI for inspecting weights, finding unhealthy tensors, and comparing checkpoints, from hand-written binary parsers.

Rust / Memory mapping / SIMD
Source ↗
05 / CONTINUAL LEARNINGScreening studies reported
Study 1 · figure from the project’s research paper

Domain-Shift Forgetting

What does a model leave behind?

Two prespecified studies of normalization and forgetting after a training-domain switch. No predicted excess forgetting established.

PyTorch / Paired seeds / Controlled experiments
Source ↗
06 / ARTIFICIAL LIFEIn progress
Ecosystem prototype · local wetland view

Artificial Life

Memory, competition, and life at the water’s edge.

A browser-based artificial-life lab exploring foraging, spatial memory, and resource competition, with a 3D ecosystem you can inspect creature by creature.

Rust / WebAssembly / Three.js
Not yet on GitHub
ALWAYS BUILDING. ALWAYS ADDING.THE COLLECTION CONTINUES ↗
02 / THE RESEARCH NOTEBOOK

The next
good question.

OPEN INQUIRY

Ideas worth testing. Each proposal starts with a question and ends with a way to measure progress.

R.01
AGENT RELIABILITY

When should an agent abstain?

Measure the tradeoff between useful action and unsupported confidence.

Proposed

The experiment

Extend OpsPilot with controlled missing, stale, and contradictory evidence. Compare always-answer, rule-based abstention, and confidence-threshold policies on the same reviewed cases.

What would count as progress

Report unsupported-claim rate, useful completion rate, escalation frequency, and latency across evidence conditions. Publish the cases and failure taxonomy alongside the curves.

R.02
REPRESENTATION & REASONING

Same problem. Different coordinates.

Does a model’s spatial reasoning survive a change in representation?

Proposed

The experiment

Build on Shadow Twins’ planned symmetry comparisons. Rotate equivalent objects, permute coordinate descriptions, and compare answers while holding the certified optimum fixed.

What would count as progress

Measure paired score differences and invalid-answer categories. Keep variants clustered with the original instance; distinguish representation sensitivity from task difficulty.

R.03
CONTINUAL LEARNING

Can forgetting be caught early?

Explore whether early diagnostics anticipate persistent deterioration.

Proposed

The experiment

Following the completed screening studies, design a separate exploratory study of checkpoint-level loss and representation diagnostics around a domain switch. Keep the existing pilot protocol unchanged.

What would count as progress

Test prespecified early-warning signals on held-out runs. Compare against simple loss-based baselines and report false alarms as well as detection lead time.

03 / BEHIND THE WORKENGINEER BY PRACTICE.
RESEARCHER BY CURIOSITY.

The interesting part
is underneath.

I’m Daniel Shdeed, a Full Stack Engineer building toward AI engineering through hands-on systems work and independent study.

I’m drawn to the space between a model that produces an answer and a system that deserves to be trusted. The data path. The failure mode. The experiment that challenges the assumption.

01

Build it.

Interfaces, typed APIs, workers, and the infrastructure that connects them.

02

Measure it.

Exact scoring, paired comparisons, reproducible runs, and honest limits.

03

Question it.

Look inside the model. Follow the evidence. Make the next experiment better.

A GOOD CONVERSATION IS A GOOD START.
What’s next?

AI engineering. Thoughtful systems.
Research-minded teams.

Email ↗LinkedIn ↗GitHub ↗
PROJECT / FIELD NOTES