How I build
My marketing runs on a personal knowledge system and the agents around it. They watch, draft and file. People decide.
The system
HoverTap any block to see what it does and how a team could adopt it.
Agents read key channels every hour and pull out decisions, asks and blockers into the briefing.
For a teamStart read-only on 3–5 channels; let the team see the digest before any agent posts.
Every four hours: triage, thread summaries and draft replies waiting for approval.
For a teamBegin with a shared inbox. Drafts only, never auto-send.
Syncs are recorded and transcribed, then turned into owners, deadlines and follow-ups on Slack.
For a teamPilot on one recurring meeting and compare against human minutes for two weeks.
Transcripts, films and old decks become raw material: one source, many formats.
For a teamDrop existing assets into Resources and tag them by topic; the archive becomes a content bank.
One knowledge base: Projects (live), Areas (ongoing), Resources (reference), Archive (done). Every input lands here, so agents always have context.
For a teamAgree one folder structure and naming convention; that alone makes the team searchable.
Scheduled agents with narrow briefs: monitor, repurpose, convert meetings to actions. I design and route them.
For a teamOne agent, one job, one human reviewer. Add the next only when the first is trusted.
Morning briefings, content drafts and action items. Nothing ships without a yes, and every output is filed back into PARA, so the next run knows more.
For a teamUse an approval queue in Slack; approved work is archived automatically.
Five principles
- Always-on monitoringSlack hourly, email every four hours, a daily sweep.
- One source, many outputsOne transcript becomes posts, ads, blogs and reels.
- Meetings become actionEvery sync ends as owners and deadlines on Slack.
- Architect, not operatorI design briefs and routes; agents do the repetition.
- People approveNothing ships without a human yes, logged back in.
An audio processor, in hardware
University of Toronto, built with Rahul Verma.
Playing through the processor. The video loads from YouTube when you press play.
A guitar goes in, the board processes it in real time, and it comes out of the speakers. We programmed it in assembly on an Altera DE2 board.
- Board
- Altera DE2, Cyclone II FPGA
- Toolchain
- Quartus and Nios II, written in assembly
- Design
- Interrupt-driven; finite-state machines run the modes
- I/O
- Audio codec in and out; on-screen menu over VGA
- Effects
- Clean, overdrive with or without feedback, reverb, mono left or right
- Transport
- Record, pause and play back
Riff: ‘A Flock of Birds’ by Karan
Other engineering work
- Forecasting energy demandAmity University · final-year projectA machine-learning model that forecasts electricity demand.
- A storage server in C++University of Toronto · courseworkA client–server storage system written in C++.
- Digital Control and Protection of Electrical MachineConference paper · 2016Published during the B.Tech in Electrical Engineering.