Backend · Cloud · AI engineering · Toronto

Your idea, running in production.

Vertex Group turns ideas, prototypes and manual work into secure, reliable software that teams and customers use every day. Web apps, backends, cloud and AI, for businesses of any size.

Taking on 1–2 new clients · Toronto and remote
{ } AI API
10+production Kubernetes clusters shipped to with GitOps
500K+events a day watched by a monitor I built
65%lower lookup latency after a database redesign
0.6 son-device AI transcription in my macOS app
Services

What I build for you

You know your business and your customers. I handle the part that makes software dependable: servers, data, security and the AI plumbing.

AI into production

You have a prototype, a script or a ChatGPT workflow that works. I turn it into a product your staff and clients can rely on.

  • Document and drawing reading, cited answers (RAG)
  • Logins and permissions so your AI stays private
  • Cost limits, caching and monitoring on every AI call

Backend and APIs

The services behind your website and apps: built to handle many customers, fail safely and be easy to change later.

  • APIs in Python, Go, Java or Node
  • Postgres, Redis, Kafka and Cassandra
  • Payments, quotes and customer portals

Cloud hosting and DevOps

Hosting that's secure, monitored and repeatable, in a Canadian region when your clients or government work require it.

  • AWS, Kubernetes, Docker, Render
  • Terraform and GitOps deployments
  • Alerts and dashboards so you hear about problems first

Automation and internal tools

Replace the spreadsheets and copy-paste that slow your team down with small tools that just work.

  • Estimating and quoting tools
  • Dashboards over your own data
  • Integrations between the apps you already use
Selected work

Problems solved, with the numbers

6issues caught on one sample plan

AI drawing review and takeoff

Prototype · construction
Problem
Contractors read every drawing by hand to catch mistakes and measure quantities before they can quote. Anything missed becomes an expensive surprise on site.
Approach
A pipeline that reads each sheet, finds the scale, measures areas from the drawing's own lines, checks them against code and the firm's standards, and lets an estimator approve every flag before a quote is built.
Result
On a sample hardscape plan: 6 issues flagged with fixes, 1,194 sq ft measured, and a full quote built in seconds.
ClaudeRAGComputer visionPython
A blueprint of a backyard hardscape plan with red revision clouds marking six issues the AI found
10+production clusters

Multi-tenant enterprise SaaS platform

Day job · platform engineer
Problem
An enterprise product had to run separately and securely for many customers, across many Kubernetes and OpenShift clusters, without a person babysitting each one.
Approach
Go and Python services for each customer's lifecycle and sign-in (OIDC), Helm and Terraform for every environment, GitOps pipelines for releases, and a health dashboard for the on-call team.
Result
Releases go to 10+ production clusters the same way every time, and the team sees every customer's health in one place.
GoPythonKubernetesTerraformGitOps
Git repoevery change reviewed GitOps pipelineHelm · Terraform cluster 1 cluster 2 cluster 3 … 10+ in production Health viewevery customer
65%faster lookups

Event pipeline for ESG reporting

Previous role · backend engineer
Problem
Reporting services built on Kafka and Cassandra were slow to look up data, and failures could go unnoticed.
Approach
Redesigned the Cassandra schema around how data was actually read, and built a monitor that watches the event stream around the clock.
Result
65% lower lookup latency, and more than 500,000 events a day under watch.
JavaSpring BootKafkaCassandra
ServicesESG data in Kafka500K+ events a day ConsumersSpring Boot Cassandra65% faster reads watches 24/7 Event monitor
0.6 sto transcribe 7 s of speech

Utter: private AI on the device

Own product · macOS
Problem
Voice dictation tools send your audio to the cloud. I wanted fast, polished dictation that never leaves the machine.
Approach
Whisper running locally on Apple's MLX, Apple's on-device model to clean up the text, and a guard that throws out any output where the AI answered instead of cleaning.
Result
A 3 MB app that types polished text into any app in 0.6 seconds, with no audio leaving the Mac.
SwiftWhisperMLXOn-device AI
nothing leaves the Mac Hold a keyand speak Whispertranscribes Apple modelcleans the text Your appin 0.6 s
How we work together

Start small, then build what works

Every project starts with a short, fixed-price discovery, so you know the plan and the price before committing to a build.

  1. Discovery

    1 week · fixed price

    I review what you have, talk to the people who'll use it, and give you a written plan with a fixed price for the build.

  2. Pilot

    4–8 weeks · fixed price

    A working version your team uses on real jobs, delivered in weekly milestones you can see and test.

  3. Run and grow

    monthly support

    Hosting, monitoring, fixes and new features, with a clear monthly scope and no surprises.

About

Hi, I'm Rajat.

I'm a backend and platform engineer in Toronto, and Vertex Group is my registered Ontario business. By day I build a multi-tenant enterprise SaaS platform that runs on Kubernetes for many customers. Before that I built Java microservices on Kafka and Cassandra for ESG reporting.

I care about systems that fail safely, software that's hard to misuse, and numbers that prove a change worked. Outside work I build things to understand them: a database engine, a demo bank, a private voice app for the Mac, and a browser game.

I'm a self-starter who learns fast and takes ownership. When something breaks, I fix the cause, not just the symptom.

Based in
Toronto, Ontario · working with clients across Canada
Education
Master of Applied Computer Science, Concordia University (Golden Key Scholar) · B.Tech in Information Technology
Languages
Go, Python, Java, TypeScript, Swift
Platforms
Kubernetes, OpenShift, AWS, Terraform, Postgres, Kafka, Redis
AI
Claude and OpenAI APIs, RAG, Whisper, on-device models
Elsewhere
LinkedIn · GitHub · Portfolio
Questions

Before we start

Who owns the code?

You do. Everything I build for you, including the code, designs and data, belongs to you once it's paid for. I keep only my general tools and know-how.

Will you sign an NDA?

Yes, before we get into details if you'd like.

How will I know how it's going?

A short update every week, a working version you can try at each milestone, and one place to ask questions any time.

How does payment work?

Discovery is a fixed price. Builds are split into milestones: a deposit to start, then a payment as each milestone is delivered. You get proper invoices from a registered Ontario business.

Can you work with my current developer or team?

Yes. I can take one piece, like the hosting or the AI, and fit into how your team already works.

I only have an idea. Is that enough?

Yes. That's what discovery is for: we turn the idea into a plan, a scope and a price before anything gets built.

Do you take small jobs?

Yes. Fixes, integrations and small tools are all welcome.

Contact

Have something you want built?

Tell me what you're trying to do in a few lines. I reply within one business day, and the first call is free.

Emailrajat.jc@gmail.com
LinkedInlinkedin.com/in/rajatjc
HoursWeekdays after 3 PM ET, or by appointment