Ian Bigford

About

I’m Ian Bigford - I build AI models, systems and products with a focus on building safe, trustworthy and useful agents. The ability for computers to improve from feedback will never cease to amaze me and I think it has the potential to make work more meaningful, and drive productivity gains that enable us to do things we couldn’t before.

Background & Inspiration

My path into computer science and AI began with studying philosophy. Specifically formal logic, philosophy of mind, and epistemology. I was fascinated by the way that formal logic - the centuries old attempt to make language into mathematically sound deductive propositions, is the basis of all programming languages and has a deep connection to rationalism. The recent advances in Generative AI has introduced a fundamental shift in the underlying epistemology of the field, introducing induction as a core component of how ai-native products will perform in production.

Quick facts

  • Education: BA Philosophy (University of Guelph); MBA (Wilfrid Laurier University); MSc Computer Science (Georgia Tech)
  • Experience: 9+ years
  • Location: Canada

Areas of Deep Work

  • Agentic systems: Multi‑agent orchestration (Plansearch, R*), small‑language model fine tuning with RL and SFT, small language model systems as sub‑agents and tools, deep research agents, two‑tower retrieval and ranking architectures, production RAG pipelines, Google ADK, and MCP‑based tool integration.
  • Computer Vision: World models (VJEPAs), mutlimodal LLMs, Vision Transformers, CNNs & more.
  • Audio Biometrics: Perception pipelines, SSL post-training, deepfake detection.

Flagship: AI-SPY

Role: CTO & Co‑founder.An AI Speech Detection Platform & Mobile App featured on CBS News nationally with over 10k users. See the website, the web client, and the mobile client.

Approach

  • Started with classic CNN baselines; evolved toward modern transformer‑based vision and stronger data/evaluation pipelines.
  • Built agentic perception/control loops and a clean feedback system to close the gap between user experience and model behavior.

Impact

  • Raised a successful seed round at a $5M valuation.
  • Secured a contract with a top global news organization for speech detection services.
  • Reached 2,000 monthly active users on the mobile app.
  • Launched the free‑to‑use web version; 500 users in week one with no marketing.

Selected work

  • HR Policy Agent (Geotab) — Built an ingestion pipeline and RAG‑powered agent that answers employee questions about company policies. Non‑technical HR staff update source documents in Google Drive, which sync to a cloud bucket and are automatically encoded for retrieval. The agent also generates employment letters on demand.
  • Sales Intelligence Agent (Geotab) — Designed a system that builds rich profiles for every sales opportunity using a deep research agent (Google ADK) orchestrated by a weekly cron job. The agent scrapes key contacts from email, ingests call transcripts, and pulls SEC filings to understand industry dynamics and deal positioning. A two‑tower architecture matches an embedding that compresses all opportunity context against a second embedding trained on historical conversion patterns to predict which deals are most likely to close throughout the sales cycle.
  • Agentic AI Knowledge Base Platform — Next.js + FastAPI + multi‑agent orchestration with streaming responses and intelligent knowledge synthesis. See the project.
  • ScriptureChat (iOS) — a retrieval‑backed companion that emphasizes clarity and citations. See the project.
  • Flourish (iOS) — therapy companion built with a reliability‑first posture. See the project.

How I work

  • Product first: start from the user problem and ship the smallest valuable thing quickly.
  • Small, fast loops: tight iteration cycles across data, models, and UX with instrumentation and evals.
  • Reliability by design: clear contracts, streaming where it matters, observability, and graceful failure modes.
  • Collaborative: clear docs, principled trade‑offs, and maintainable systems.

Research & writing

I’m beginning to publish practical notes on agentic systems and applied AI. Stay tuned and check the blog for new posts.

Stack & tools

  • Frontend: Next.js (App Router), React, TypeScript, Tailwind, Framer Motion
  • Mobile: React Native
  • Backend: FastAPI (Python), SQLite/aiosqlite, server‑sent events
  • Orchestration: LangGraph, ADK, MCP; integrations with external data sources
  • ML/Data: PyTorch, NumPy, Pandas
  • Cloud: GCP/AWS

Ethics & responsible AI

I focus on transparent evaluation, user‑centered design, and safe defaults. In AI-SPY, we couple perception models with clear feedback loops and guardrails so the system remains useful, reliable, and respectful of users.

What’s next

I’m not consulting. I’m exploring roles at top engineering organizations where I can build applied AI systems—agent platforms, perception, and real‑time products—at scale.

Contact & links