I'm Benedict Anokye-Davies, a University of Nottingham Computer Science with Artificial Intelligence student graduating in June 2028. I build Python trading software for a client and contribute to team projects in simulation, research and model evaluation.
Selected projects and client work, with the problem, contribution and evidence stated plainly.
Trading workflows must avoid mistaking a restart or partial process for a confirmed outcome.
I develop Python, FastAPI and PostgreSQL features for persistent order tracking and restart recovery that checks exchange fills.
Dashboard caching is covered by tests, alongside recovery checks that require exchange evidence before success is recorded.
Scheduled financial-event research needs a workflow the team can repeat and inspect without turning it into an autonomous trading claim.
I contributed to event research and paper-trading tooling, including MCP order actions, with deterministic controls around structured LLM reasoning.
The team exercised a clean-installed package in an offline four-event evaluation. Orders are paper-only.
Evaluate environmental-control ideas in repeatable scenarios before relying on a learned component.
I helped build a connected-zone simulation and model-evaluation work for airflow, temperature, resource use and injected faults.
On the held-out fault-detection evaluation, the rule baseline beat the learned detector. Plant actions remained deterministic.
Archived imageMake simple ecological rules easier to inspect through an interactive browser model.
A personal ecosystem experiment built around procedural terrain, inherited traits and environmental pressure.
The browser demo is exploratory rather than a validated ecological predictor.
I like systems that make their assumptions inspectable: persistent state instead of hopeful recovery, deterministic controls around model reasoning, and tests that say what a change is meant to protect.