Research
Research across reinforcement learning, language models and agentic systems.
- Reinforcement learning
- Reward modelling
- Simulation and environments
- Agentic systems
- Language models
- Fine-tuning and alignment
AI
API
UI
AI
- def project_observation(project, schema):
- risk = np.array([
- project.tasks.ratio_with_risk(c)
- for c in schema.risk_categories
- ])
- location = one_hot(project.location, schema.locations)
- building = one_hot(project.type, schema.building_types)
- scale = np.array([
- min(project.units / schema.max_units, 1.0),
- min(project.storeys / schema.max_storeys, 1.0),
- ])
- return np.concatenate([risk, location, building, scale])
API
- async def retry_with_backoff(call, max_attempts=5, base=0.5, cap=30):
- last_error = None
- for attempt in range(max_attempts):
- try:
- return await call()
- except RetryableError as error:
- last_error = error
- if attempt == max_attempts - 1:
- break
- delay = min(cap, base * 2 ** attempt)
- delay *= 1 + random.random() * 0.25
- await asyncio.sleep(delay)
- raise last_error
UI
- export const load = async () => {
- const res = await fetch("/api/scenarios");
- return { scenarios: await res.json() };
- };
- export const actions = {
- create: async ({ request }) => {
- const data = await request.formData();
- await fetch("/api/scenarios", {
- method: "POST",
- body: JSON.stringify(Object.fromEntries(data)),
- });
- }
- };
Projects
Projects span research, platforms and production systems, from first-principles experiments through to systems running at scale.
AI platforms and products
Full-stack engineering for AI-driven platforms and products, covering backend services and production systems
Research infrastructure
Training pipelines, experiment orchestration, evaluation harnesses and research-to-production tooling
Address
167–169 Great Portland Street5th Floor, London W1W 5PF