TD Securities · COO · Internal

TD Builders Club

Stop watching AI happen. Get your hands ON!▋

8 sessions · passion projects · the newest secret sauce · snacks included

Open warm. "This is the club that gets the rest of us building." Don't over-explain — the next 13 slides do that. End on the tagline.

The problem

Smart people. Sidelined by AI.

01 · AI anxiety

Everyone feels left behind.

The pressure to be "tapped in" is relentless — and it doesn't go away just from watching demos or sitting through lunch-and-learns. People want to build, but there's no structured on-ramp that feels worth their time.

02 · The gap

Tutorials don't translate.

Plenty of people have done a Python course or poked at ChatGPT. The jump from that to building something real — something that actually runs, does something useful, and you made — requires reps that no tutorial gives you.

03 · FTEs especially

No forcing function to build.

Interns have projects. Engineers have tickets. FTEs have neither — no hackathon invite, no assignment, no sandbox time. Without a reason to build, the default is to not build at all. This club fixes that.

Three pain points, one breath each. Don't dwell — the rest of the deck is the answer. Land the FTE beat hardest — that's the audience that needs this most.

What it is

A skill. A build. A demo.

01

7–10 minute skill presentation. see slides 6–13

02

40 minutes of building. mentors float

03

DEMO TIME! show off what you made

Come have fun. Finally test that idea of yours.▋

Counter the corporate-training instinct. This isn't training. It's a club. The "finally test that idea of yours" line is the hook — say it like you mean it.

How a session runs

Run of show.

01 · Arrival

Doors + floor open.

Build while you arrive. Snacks. Music.

15 min early

02 · Skill drop

The 7-minute talk.

One sharp idea. One demo. Then we shut up.

7–8 min

03 · Build

Build time.

Execs float. Bring your own thing or pair up.

~45 min

04 · Show

Optional demos.

Show off. Get feedback. Round of applause.

last 10 min

⟶ Virtual attendees in breakout rooms — same shape, same energy.

The cadence is the product. Talk → Build → Demo. Mention virtual attendees explicitly — half the room is probably watching from Toronto/NYC.

The season

8 weeks. One arc.

S01 · June 24th

Prompt Monkeys

Prompt engineering skills and context management, serving as a foundation for future sessions. Will draw into AI safety and Copilot quick-setup too.

S02 · July 8th

Eye Candy

UI/UX component libraries and resources, such as 21st.dev, Figma Make, and Hero UI. Design thinking for AI Systems, to fix up vibecoded apps.

S03 · July 15th

API Haven

Throw attendees some super cool APIs, such as ElevenLabs for voice agents or a computer vision body language detector — endless possibilities.

S04 · July 22nd

Lies of Copilot

This is the vegetables session. Quick look into AI safety in terms of data, privacy, and bias. (Removed if 7 weeks)

S05 · July 29th

AI at Home

Dive into AI, learning the fundamentals of fine-tuning, RAG, and ML. Important for developing your own AI SaaS!

S06 · Aug 5th

Build an Intern

Self-explanatory — agentic workflows.

S07 · Aug 12th

Fullstack Toolkit

Tools for building quickly, such as Supabase, Vercel, and Lovable.

S08 · Aug 19th

Mini-Hackathon

2 hours, demos, prizes.

foundation→ make it pretty→ plug things in→ safety→ under the hood→ make it yours→ ship it→ hackathon

The arc strip is the takeaway: each session feeds the next. By S08 they have the whole stack. Point at S03 (API Haven) and S07 (Fullstack) as the highest-leverage weeks.

Session 01 — Prompt Monkeys

S01

June 24th

Prompt Monkeys.

Stop prompting like it's Google.▋

You'll learn: Context + structure beats clever wording. How to give a model enough to actually do what you mean — not what you typed.
Tools: KMAI sandbox Internal chat model
The hook: Live prompt makeover — same task, lazy vs. engineered. The difference is embarrassingly large.
Why it's engaging: Everyone secretly suspects they're doing it wrong. Before/after payoff is immediate. Sets the tone for the whole season.

The opener. Set the tone for the whole season: this is hands-on, slightly cheeky, grounded in real work. The 7-min talk is the standalone Prompt Monkeys deck.

Session 02 — Eye Candy

S02

July 8th

Eye Candy.

Look good without being "creative."▋

You'll learn: How to pull from ready-made component libraries to skip the blank-canvas problem entirely. Design thinking for AI systems — fixing the vibe-coded mess after the model spits it out.
Tools: Figma Make 21st.dev Hero UI
The hook: Blank screen → polished, usable UI inside one session. No design background needed.
Why it's engaging: Directly addresses the "I'm not a creative person" barrier. People leave with something that looks like a real product — not a student project.

Aim this at the analysts who "don't think they have an eye." Show them they don't need one.

Session 03 — API Haven

S03

July 15th

API Haven.

An API is just a digital plug.▋

You'll learn: How to borrow capabilities you'd never build yourself — voice, vision, real-time data — through a single call. The ceiling goes way up.
Tools: ElevenLabs Computer vision APIs Streamlit
The hook: Voice agents, body language detectors, live dashboards from a photo. Pick one and wire it up in the session.
Why it's engaging: The "wait, I can just plug that in?" moment. Opens doors people didn't know existed — and it's different for every attendee depending on what they build.

idea → find the API → plug it in → ship

The flow strip is the teaching beat: "you didn't build the eyes. You borrowed them through a plug. That's an API." Backup demo (if vision endpoint isn't exposed): Earnings Brief — text in, brief out.

Session 04 — Lies of Copilot

S04

July 22nd

Lies of Copilot.

Catch the AI lying. Live.▋

You'll learn: How to spot confident-but-wrong answers, where data leaks happen, and what safe sandbox use actually looks like day-to-day — not as a policy, but as a habit.
Tools: KMAI sandbox Mostly conceptual
The hook: Catch the model confidently wrong about something TD-specific — live in the room. Then talk about why that matters.
Why it's engaging: Framed as a survival skill, not a compliance lecture. People lean in when you show them the AI is fallible — it reframes the whole relationship with the tool.
Note: Will be removed if we only have 7 weeks instead of 8.

The "vegetables" session — but framed as a survival skill, not a lecture. Pre-stage two known-bad prompts in case the room is shy. Note: This session will be removed if we only have 7 weeks instead of 8.

Session 05 — We Have AI at Home

S05

July 29th

We Have AI at Home.

LLM vs. fine-tune vs. RAG — finally.▋

You'll learn: The actual difference between a base LLM, a fine-tuned model, and a RAG setup — in plain English. Foundational knowledge for anyone serious about building their own AI product.
Tools: Internal models RAG on internal data
The hook: "Why doesn't the generic bot know our deal flow?" Because nobody gave it the docs to read. Here's how you fix that.
Why it's engaging: Demystifies jargon people have been nodding along to for two years. By the end, people can make real decisions about which approach fits their use case.

Plant the RAG seed in S01 so this session pays it off. The three boxes (LLM / fine-tune / RAG) need plain-English one-liners not architecture diagrams.

Session 06 — Build an Intern

S06

Aug 5th

Build an Intern.

Work while you sleep.▋

You'll learn: How to set up agentic workflows that run on a schedule, watch for triggers, and take action without you in the loop. The basics of autonomous AI.
Tools: Power Automate Copilot Studio Google Agent Development Kit
The hook: An agent that watches a spreadsheet or inbox and acts on it overnight. You wake up and the work is done.
Why it's engaging: The pitch writes itself for anyone with a recurring task they hate. No-code entry point means everyone in the room can leave with something that actually runs.

No-code agents — the lowest barrier of the whole season. Anyone can leave with a working intern.

Session 07 — Fullstack Toolkit

S07

Aug 12th

Fullstack Toolkit.

A real URL. A real database. Yours.▋

You'll learn: How to take a frontend and turn it into a full application — one that stores data, lives on a real URL, and doesn't disappear when you close the tab.
Tools: Vercel Supabase Lovable
The hook: Deploy something with a real URL and a working database inside the session. It exists on the internet. You made it.
Why it's engaging: "My app is actually real" is the moment the whole season has been building to. Arms everyone for the finale. The gap between prototype and product disappears.

The Holy Grail. "My app is actually real" is the moment the whole season hinges on. Pre-build the demo URL — don't deploy live on stage.

Season Finale — Mini-Hackathon

S08

Aug 19th · Finale

The Mini-Hackathon.

Bring it. Ship it. Demo it.▋

The shape: 2 hours. AI sandbox open. Snacks. High energy. Everything from the season in one room.
Bring: Whatever you've been building all season — dust it off, ship it, show it.
End on: Demos, prizes, round of applause, and a photo. A real finish line.
Why it's engaging: People work harder when there's something to show for it. Prizes and demos create the kind of energy that makes people want to come back next season.

Build energy into how you say this slide. This is the payoff. Confirm snacks + a couple of small prizes if budget allows.

Closing

Come build something.
See you at the workbench.▋

Wednesdays · 11am · Aspen Room · Floor 19

Join → #td-builders-club on Teams

Standing rule · Build with synthetic or public data only. Never real TD data in external tools.

Loud on the standing rule — this is the one thing they have to remember. Don't bury it. End with the CTA — channel, time, place. Open the floor.