Arcanist

Engineering capacity,
on demand.

You describe what you need in Slack. The agent does the work, ships a PR, self-heals through CI, and hands back code ready to merge. No hiring. No onboarding. No waiting.

tryarcanist.com

01 - Who We Are

We're Shivam and Josiah.

Met on the bus home from the first day of 9th grade. Both University of Waterloo. Both technical. Both builders.

We've been building together for over a decade - first as friends, then as engineers, now as founders. We don't just want to build a company. We want to win.

Our friends stopped playing pickup basketball with us because we're too competitive. That's how we build too.

02 - Founder

Shivam Pandey

Employee #1 at Raylu (Series A). Didn't join to learn - joined to build every core AI system from research to production, solo. 0 to $XM ARR in under a year.
Built the first-ever ML solution for proactively preventing phishing attacks at RBC Global Cybersecurity. Patented. NLP and Computer Vision model safeguarding banking credentials and preventing up to C$15M in losses per year.
Helped close multiple funds at PE firms by being in the sales process end-to-end - ensuring AI systems deliver real value, working with technical and non-technical stakeholders, getting deals over the line. Knows how to handle the buyer, the build, and every angle in between.

Most founders can build or sell. I do both - and I've done both at scale before Arcanist existed.

03 - Founder

Josiah Parappally

#1 in merged PRs at Ramp over the last 12 months - by over 500. #1 over 24 months by over 700. #2 all-time. Fastest intern-to-Staff promotion in Ramp history. Not reputation - measurable, undeniable output.
Tech led multiple P0 product lines to GA - Excel plugin, reporting, AI reporting, budgeting. Grew a direct report from B2B to Senior in under a year. Now TL of Ramp's new construction vertical. 0-to-1 products are what I do.
Built Ramp for Excel on nights and weekends on pure conviction. 500+ businesses. Fastest-growing Ramp integration of 2024 by 2.5x. 300x average engineer usage of Ramp's internal coding agent. Instead of staying to lead the next thing at Ramp - I'm here. Because this is bigger.

I don't wait for permission to build. I've never needed someone to tell me what needs to exist.

04 - The End Vision

Solve the codebase,
unlock every team.

Once you have an agent that truly understands a codebase, you unlock far more than engineering tasks. Every function in the company touches software - and the agent can navigate it.

Customer Support

Stop pinging engineers

CX agents query the codebase directly in read-only mode to answer technical questions - no tickets, no waiting.

Product Marketing

Read the source

PMMs read the codebase to build accurate technical material, changelogs, and positioning - without bothering the team.

Operations

Self-serve answers

Ops pulls answers from the system directly instead of filing a ticket and waiting three days.

05 - Today

Tag @arcanist in Slack.
Ship by end of day.

No new IDE, no new platform, no new tab. Describe what you need in plain English, right where your team already works.
Full-stack context from day one. GitHub, Sentry, Datadog, CI/CD, databases - the agent understands your stack.
Self-healing. Creates PRs, reads CI logs, fixes failures, re-pushes. You review the final result, not the process.

# engineering

SP

Shivam Pandey 2:14 PM

@arcanist checkout flow is throwing 500s on /payments. Can you pull the Sentry errors, find the root cause, and open a PR with the fix?

A

Arcanist APP 2:14 PM

On it. Pulling Sentry errors for /payments...

Found it - null pointer in stripe_handler.py:142. Opening PR now.

A

Arcanist APP 2:17 PM

PR #847 ready for review. CI passing.

Fix null check in payment processing

06 - The Gap

Every AI tool solves for code execution.
That's not the hard part.

The hardest 20% of what makes an engineer great is outside the code itself - understanding how your CI works, your team's conventions, your architecture decisions, your deploy process, the tribal knowledge that lives in someone's head.

That 20% is different for every company. And nobody is solving for it.

Existing tools generate code. We generate capacity - work that's actually shippable in your specific environment.

07 - Our Approach

We solve the 20% that's
different for every company.

Not smarter code generation - deeper integration with how your specific company operates
The agent learns your CI pipeline, your naming conventions, your deploy process, your architecture
Every deployment teaches us more about how companies actually work - that compounds

We're the last mile of AI engineering. Every tool can generate code. We deliver work that actually ships in your specific environment.

08 - The Abstraction Ladder

We always move up a layer
before everyone else.

Our advantage has always been operating at a higher level of abstraction than our peers. When they catch up to where we are, we've already moved up again.

We believe a swarm of background agents is the final level. Not code generation - full engineering capacity. Context, CI, deployment, iteration, self-healing. Agents that own the lifecycle end to end.

We're going to evangelize this to everyone. This is the endgame of software engineering.

Our competitors are building better code generators. We're building the layer above - where agents don't write code for you, they do the work for you. That's the gap that compounds.

09 - Who We Serve

The coding tool for
the understaffed.

Series A through C companies. Too much complexity, not enough time. The team where one engineer owns three services and the backlog is six months deep.

They don't need another autocomplete. They need capacity - work that gets done without hiring, onboarding, and waiting three months for someone to ramp.

These are the teams where the stakes are highest and existing AI tools fail hardest - because the codebase is complex, the systems are interconnected, and the context can't be faked.

The ICP

High complexity,
low manpower

Series A–C · Complex codebases · Lean engineering teams · Growing faster than they can hire

10 - Market & Business Model

Hiring budgets are becoming
tooling budgets.

Companies aren't going to hire their way out of the complexity problem anymore. That budget is moving to tooling - and we're where it lands.

Per-seat pricing for the core product. Usage-based cross-sell as teams route more work through the agent. Strongest position: indispensable per-seat tool that naturally expands into consumption.

The customer type that attaches fastest: scaling startups with significant engineering complexity. The team where one engineer owns three services, the backlog is six months deep, and every AI tool they've tried generates code that doesn't ship.

$42B Developer tools market

$250K+ Avg. fully loaded SWE cost (the budget we displace)

11 - Go-to-Market

Bottleneck is support,
not demand.

First design partner deployment live within 2 weeks of funding. Full launch in 2 months. Once the product is solidified through design partner work, we blaze to $1M ARR in a matter of months.

The constraint isn't finding customers - it's our ability to support them. Demand is already ahead of us. The network gets us in the door; the product keeps us there.

Multiple companies competing to be design partner #1
Path to $1M ARR on network alone
Product expands organically within teams once deployed

GTM Edge

We know how
engineering buys

Josiah led procurement of multiple engineering vendors at Ramp - knows exactly how engineering teams buy. Shivam sold to in-house technical teams at PE firms - knows how to navigate non-obvious buyers. Both know how to find the champion, prove value fast, and expand.

Use of Funding

Everything goes into
building and shipping

Core engineer hire. Legal and infrastructure foundation. Both founders full-time. Cloud compute for production deployments. Everything goes directly into building and shipping.

12 - Competitive Landscape

The expected threats
aren't threats.

Claude Code

Great for individual developers writing code faster. But it doesn't know your CI pipeline, your deploy process, your team's conventions, or your architecture. It generates code - it doesn't generate capacity.

Cognition (Devin)

Impressive demos, focused on autonomous coding. Same fundamental gap - it doesn't integrate with your Slack, your Sentry, your specific environment. It works in isolation. Real engineering doesn't.

Our Moat

The last mile
compounds

We're not competing on code generation - every model gets better at that. We're competing on the last mile: deep integration with each customer's specific environment. CI pipelines, naming conventions, deploy processes, architecture decisions, tribal knowledge. That learning compounds with every deployment. It's not something you can shortcut.

13 - Bet On Us

It's pretty obvious
this is a good idea.

The market is clear. The timing is right. Every company needs more engineering capacity and the budget is moving from hiring to tooling. The real question isn't whether this works.

The real question is: why us?

Other people in this game are smart. But nobody will outfight us. We love the fight. Nobody else in this game is going to get punched in the face and get up with more determination than they started with. That's the bet. Not just the product - us.

Arcanist