Tai Lin

Founding Engineer · AI Infrastructure · Capital Markets

Tai Lin

I build the infrastructure layer between probabilistic models and deterministic financial systems.

Available for consulting $0 → $350K ARR ↗ 6+ agents in production ↗
Fig. 01 The layer between two opposite worlds Click to feed the layer

I find meaningful 0-to-1 problems and build the systems that make autonomous agents safe in regulated domains — the harnesses, capability registries, and audit layers that sit between a probabilistic model and a funded transaction.

Selected work

2017 — present

HighFi

Founding Software Engineer

Mar 2023 – Present

  • TypeScript
  • PostgreSQL
  • Next.js
  • Vercel AI SDK
  • Pulumi
  • GCP

$350K

ARR scaled

HighFi is agentic infrastructure for capital markets. It replaces the manual, operations-heavy back-office work that lenders and asset managers still run by hand with self-serve systems that don't need a person in the loop. As founding engineer and the primary technical contact for customers, I built the platform's core systems end-to-end and grew HighFi's first two enterprise accounts from $0 to $350K ARR.

Capital Markets OS Read →

A purpose-built operating system for the layer between a legal agreement and a funded transaction: compliance, servicing, reporting. Six components, one config registry — onboarding becomes a config task instead of a sprint.

6 components, config-driven
AI Agent Harness Read →

BaseAgent<TDeps,TInput,TOutput> powering 6+ agents with typed dependency injection, structured step tracing, and per-step Slack observability. Stateful FSM design.

6+ agents in production
Millie & Penny Read →

Capability-gated agents resolving authorized toolsets per request. Millie executes funding draws and SOFR updates in Slack; Penny streams amendment edits from a mode-scoped in-app panel.

5 capabilities, authorized per request
The Night Agent Read →

Three-stage Claude Code pipeline on GitHub Actions. Safety lives below the prompt layer: every shell command routes through a whitelist. Specs go in by day, PRs come back by 8 AM.

55m total pipeline budget

Utradea

Software Engineer → Software Engineer II

Mar 2021 – Feb 2023

Before HighFi →
  • Node.js
  • MongoDB
  • PostgreSQL
  • AWS ECS
  • Redis
  • OpenAI API

50+

APIs migrated

Utradea, now Financial Modeling Prep, is a market data platform serving free and enterprise API tiers to developers worldwide. I was the first engineering hire, with no formal CS background. I taught myself and shipped paid production features within months — Stripe payments, the SEC Dashboard, and a re-platform of the MERN backend from Lightsail onto ECS with Fargate to support 50+ production APIs. I believed in the work enough to become a shareholder.

FMP API Migration

Led the post-merger migration of 50+ Financial Modeling Prep APIs into Utradea's Node.js / MongoDB stack, owning the full infrastructure transition.

AWS Infrastructure

Re-platformed from Lightsail onto ECS with Fargate and load balancers; multi-node Redis cycle caching with subscription-tier rate limiting for high-frequency price reads.

First AI Feature

Piloted an early OpenAI API integration behind Financial Writer, a tool that generated news articles from real-time market data. Shipped before ChatGPT made the idea obvious.

SEC Dashboard & Stripe

Built the SEC Dashboard end-to-end: an NLP pipeline (winkNLP, VADER) that parsed 8-K filings into investor-readable insights. Integrated Stripe for Utradea's first paid users.

BMO Financial Group

Business Analyst → Senior Business Analyst

May 2018 – Mar 2021

Before HighFi →
  • Requirements
  • Systems Analysis
  • Regulatory Compliance

3yr

Domain depth

Three years in capital markets and commercial banking gave me something most engineers don't carry: I read the legal agreements and defined the requirements for credit systems. I understood how financial infrastructure actually works before I ever wrote a line of production code.

Vendor Risk Decommissioning

Led the decommissioning of a vendor risk management platform, owning the business requirements and the migration to a leaner in-house replacement.

Credit Flow Transformation

Defined E2E requirements and drove cross-functional delivery for the Credit Flow for Commercial Banking transformation.

Client Portal Integration

Owned E2E Client Portal Integration — streamlined business banking onboarding from prospecting to signing across legacy systems via API.

Compliance Reporting

Introduced configurable credit compliance reporting templates scaled across multiple lines of business to standardize regulatory workflows.

Creative Destruction Lab

Venture Growth in Machine Learning Stream

2017 – 2018

Before HighFi →
  • ML
  • Venture
  • University of Toronto
ML Venture Cohort

Worked alongside ML-native founders and operators during the early commercial wave of machine learning. Developed early conviction: the bottleneck is infrastructure, not capability. B.Com, University of Toronto.