Now
Founding engineer at Floe Labs
Payments and metering infrastructure for AI agents, built from an empty repo. I built the API, the console, the gateway, the voice platform, the CLI and the deploy pipeline.
Checked before money moves
The core API meters every call an agent makes and enforces its spend limit before funds move. A call over the limit stops at the check.
One gateway, 20 providers
An OpenAI-compatible LLM and voice gateway routes 125 models across 20 providers and 6 modalities.
Twenty invoices, one bill
Usage from every provider lands on one real-time bill instead of twenty vendor invoices reconciled by hand. Stripe carries it: four subscription tiers, Connect invoicing, and daily usage into Billing Meters.
Voice that stops on budget
A Node WebSocket media gateway runs a live speech-to-text, LLM and text-to-speech loop, meters each call leg, and halts a call that breaches its budget mid-conversation.
Self-serve for developers
A Next.js console, a 33-command CLI, and an MCP server with 80 tools listed in the official MCP Registry. Signup provisions embedded wallets behind email or Google login: 400 users and 746 wallets onboarded without a seed phrase.
- Read the Floe case study
Selected work
All 7 projectsFloe Labs
Founding Software Engineer, Aug 2025 to Present
The layer that meters what an AI agent spends, and stops it before it spends more than you allowed.
Quellit AI
Full-Stack & AI Engineer, 7 months
It reads the requirements a team already wrote, turns them into end-to-end tests, and runs them in a real browser.
Neyo: AI Agent Engine
AI & Full-Stack Engineer, 2 months
An agent that chains 65 tools in one conversation and streams the answer back while it works.
Docubark
Full-Stack Engineer, 12 months
Third-party risk management, minus the analyst reading every vendor questionnaire by hand.
Experience
Five years, three companies, from leading a team of five to founding engineer.
Aug 2025 to now
Floe Labs
Founding Software Engineer
Payments and metering infrastructure for AI agents, built from an empty repo. The API meters what an agent spends and enforces its limits before money moves. The gateway in front of it routes model and voice calls across 20 providers.
Show highlightsHide highlights
- users onboarded, no seed phrase
- 400
- from 20 vendor invoices
- 1 bill
- Architected the core payments and metering API in TypeScript, Node, and Postgres, scaling it to 313 REST endpoints covered by 2,193 automated tests, with spend limits enforced before funds move
- Shipped the customer-facing developer console in Next.js: 39 routes covering agent management, spend policy, billing, and webhooks, which took the product from API-only to self-serve
- Launched a walletless signup flow on Privy that provisions embedded wallets behind email and Google login, onboarding 400 users and 746 wallets without exposing a seed phrase or a private key
- Built an OpenAI-compatible LLM and voice gateway that routes and meters 125 models across 20 providers and 6 modalities, replacing manual reconciliation of 20 vendor invoices with a single real-time bill
- Engineered a real-time voice platform on a Node WebSocket media gateway, running a live speech-to-text, LLM, and text-to-speech loop that meters each call leg and halts a call breaching budget mid-conversation
- Designed a 70-table Postgres schema and migrated every backend service from SQLite to managed Cloud SQL in a single cutover, replacing 1,600 lines of ad-hoc startup code with versioned migrations
- Delivered the monetization layer on Stripe, pairing four subscription tiers with Connect-based client invoicing and daily usage reporting into Billing Meters for enterprise contracts
- Published four developer packages to npm and PyPI, including a zero-dependency CLI spanning 33 commands and a Model Context Protocol server exposing 80 LLM-callable tools, listed in the official MCP Registry
- Established the deployment pipeline for the entire production estate, automating 6 CI/CD workflows to Google Cloud with post-deploy verification that blocks a bad release instead of reporting false success
Feb 2024 to Jul 2025
Chorus One
Frontend Engineer
Frontend on the institutional staking rewards platform. Most of my time went to two things: making it fast, and keeping defective builds out of production.
Show highlightsHide highlights
- faster dashboard, 120s to 15s
- 8x
- end-to-end test coverage
- 85%
- Developed a Next.js and TypeScript platform for institutional staking rewards, normalizing data from two provider APIs and public chain RPCs across nine Cosmos networks into a single performance dashboard
- Cut dashboard load time 8x, from 120 seconds to 15, by parallelizing third-party API calls and caching USD price lookups across the GraphQL and REST provider integrations
- Raised end-to-end test coverage to 85% in Cypress and integrated the regression suite into CI, blocking defective builds before they reached production
- Pulled the table components apart into reusable pieces, which shrank the codebase and made it easier to change
Jan 2022 to Jul 2023
Rejoicehub Solutions
Software Engineer
Team lead on client web apps in React and Node.js, usually several running at the same time.
Show highlightsHide highlights
- engineers led
- 5
- release throughput
- 2x
- Led 5 engineers across client web apps, introducing code review and Jenkins CI that doubled release throughput
- Optimized the highest-traffic Node.js API and PostgreSQL query paths, cutting average response time by 30% through targeted indexing, query restructuring, and response caching
- Built reusable React components, API helpers, and the data-analytics microservices behind operational reporting
- Collaborated with design and QA to ship responsive HTML, CSS, and JavaScript interfaces across concurrent client projects, resolving cross-browser and device inconsistencies before release
Concordia University
Master of Engineering (MEng), Software Engineering
Completed Apr 2026
Uka Tarsadia University
Bachelor of Technology (B.Tech), Computer Engineering
Jul 2021
Toolbox
Pick a tool to see where I used it.
Languages
- Java
Used at
Frontend
- TanStack Query
- Radix UI
- Vitest
- Jest
Backend
- Hono
- Django
- OpenAPI
- Celery
Data
- Redis
- Drizzle ORM
- SQLAlchemy
- Alembic
- ChromaDB
AI Engineering
- Anthropic SDK
- Vector Search
- Embeddings
- Playwright browser agents
Infrastructure
- Kubernetes
- Terraform
- GitHub Actions (CI/CD)
- Vercel
- Sentry
Writing
Sep 25, 2026, 24 min read
What Is Jev? Integrating Jev With LLMs, Claude Code, and AI Agents
What Jev, TypeSafe's System One Model, is and how to integrate it with LLMs, Claude Code and AI agents, with chatbot and monitoring agent examples.
Read the post
Sep 23, 2026, 18 min read
What Is an MCP Server? How It Works and How to Build One in TypeScript
MCP explained through something you already know, an HTTP server, then a small MongoDB MCP server built in TypeScript.
Sep 20, 2026, 8 min read
What I Still Do Myself When AI Writes the Code
I use Claude Code every day and let it write a lot of my code. I still read every diff, dry run every migration, and push the code myself.
About
I'm a software engineer in Montreal, five years in. I build the backend and the AI parts of products, and the frontend that puts them in front of people.
I finished my MEng in Software Engineering at Concordia University in April 2026. Away from work you'll usually find me watching a movie, playing a video game or reading, with music playing in the background.