Skip to main content
David Toy

David Toy

Engineer and technical founder, with commercial and stakeholder judgement, who embeds to build and ship LLM-based solutions

Talent
DunblaneMember since July 21, 2026
PythonTypeScriptMongoDBAWS BedrockServerless microservicesAWSRAG (chunking, hybrid search, reranking)LangGraph orchestrationAgentic patterns (ReAct, ReWOO)Cost-aware model selectionVector search
·

Hire this person

Send a job offer directly to this candidate

PythonTypeScriptMongoDBAWS BedrockServerless microservices

About

Engineer and technical founder, with commercial and stakeholder judgement, who embeds to build and ship LLM-based solutions. Most recently I designed and built a multi-agent LLM pipeline for Appointedd, a B2B2C global bookings platform, to production spec that is currently shadow deployed against live traffic. I'm versed across the LLM stack: RAG (chunking, hybrid search, reranking), LangChain and LangGraph orchestration, agentic patterns (ReAct, ReWOO), cost-aware model selection, and vector search.

Primarily

Python, but I adapt readily to client stack. I'm comfortable owning solutions end-to-end and translating ambiguous business problems into concrete technical specs for both engineering teams and senior stakeholders.

Experience

AI Solutions Engineer at CodeClan, with Appointedd (2026-03 – Present)

Designed, built, tested and delivered, a multi-agent LLM solution that surfaces customers' "Booked Intent" to unlock additional business value for Appointedd, working embedded within the client's team to their technical and data-governance constraints.

  • Multi-stage agent pipeline as a serverless microservice, designed to handle c.500k bookings per year. Currently in shadow/dark deployment, to evaluate performance against real world conditions.
  • Cost-aware model selection: evaluated all regionally available models on AWS Bedrock (client constraint) against the use case, then mixed models by pipeline stage – a lightweight model (Nova 2) for data distillation and a highly cost-efficient model (GLM 4.5 Flash) with a detailed prompt for the final data-to-prose step – to balance quality against per-call cost.
  • Test harness and QA: wrote unit and integration tests plus a separate testing harness that hooked into each pipeline stage, enabling out-of-environment testing across customer and adversarial scenarios, and systematic prompt optimisation.
  • Adapted to the client stack: delivered in TypeScript with integration to MongoDB, moving outside my usual Python and Postgres.

Technology & Strategy Consultant at Freelance (Present)

Short- and long-form engagements spanning technical and web development, programme management, business-case development, funding applications, and product and market analysis.

Programme Manager, St Andrews Forest at University of St Andrews (2021-01 – 2025-02)

Sole lead establishing the University's nature-based carbon-offsets portfolio, taking the initiative from blank page to a signed 100-year research and sustainability partnership (c.100,000 tonnes offset, multi-million-pound value). Built trusted relationships and constructively challenged senior stakeholders across disciplines to reach agreement.

Product Manager, Marketing Manager at Cyan Forensics (2017-02 – 2020-08)

Child protection software for Law Enforcement. Ran technical marketing during GTM (face-to-face and online), then pivoted to product management once traction had been achieved.

Technical Co-founder at GeometryGeeks.bike (2016-12 – 2022-06)

Full-stack (Python, web, cloud PaaS) A side project that became a global product helping cyclists find the correct bike size by crowdsourcing validated geometry data. Successful exit via trade sale.

  • Built an ingestion-and-validation data pipeline: scraped a user-supplied page, used heuristics and statistics to locate the geometry table and normalise parameters across inconsistent manufacturer schemas, then returned the extracted data for human verification before publishing it to all users.
  • Scaled to c.100,000 monthly users on an SEO-led, near-zero CAC, freemium model.

Product Engineer (technical lead, bootstrapped team) at InclusIQ Institute (2014 – 2016)

Led a small tech team (myself, a graduate, and a part-time consultant). Made the platform choice (Google App Engine / PaaS) and coached the junior engineer, steering them toward pragmatic engineering decisions.

  • Designed the product, ran client-facing activities, secured grant funding, and edited e-learning and marketing content.

Education

Applied Agentic AI – CodeClan (2026)

Saltire Fellowship in Entrepreneurship & leadership – Babson College (2010 – 2011)

MPhil in Industrial Systems, Manufacturing & Management – University of Cambridge (2008 – 2009)

MEng in Electrical & Mechanical Engineering – University of Strathclyde (2002 – 2007)

Skills

Engineer and technical founderwith commercial and stakeholder judgementwho embeds to build and ship LLM-based solutions

Reviews

Related