the whole wall

What I've built, and what it did

What each product does for the people using it, what I built, and a simple picture of how it works. Kindred products were built with a team, and I say which parts are mine. Soonbuilt products are mine end to end.

Kindred

Senior full-stack engineer since 2023. Kindred Workplace and Kindred Practice are the products I've spent most of my time on; Kindred Core and Agent Ops are the shared AI and tooling underneath them.

Kindred Workplace: six meetings because no one could have one honest conversation
Kindred · emotional intelligence at work Measure how a team works together. Then help everyone get better at it.

Kindred Workplace

Current

Senior engineer · AI coach, conversation analysis, integrations, production

The costliest problems at work live between people: the six meetings that replace one honest conversation. Kindred Workplace measures emotional intelligence across a company and coaches every person to grow it, in the app, by voice, and inside Slack and Teams.

  • 8,300+ people on the platform
  • 21 companies using it
  • 2,400+ people with an EQ profile

What I built

  • Analyze, from scratch. Upload or record a real conversation and see how you showed up, speaker by speaker, mapped to your EQ profile. Handles recordings up to 90 minutes.
  • A coach people can trust. Made Paige, the AI coach, remember context and send useful nudges, and put quality checks and failure monitoring behind every release.
  • Coaching where people already work. Paige inside Slack and Microsoft Teams, so growth happens in the flow of work, not in another tab.
  • Growth for the business. Partner referral codes and commissions, adoption dashboards for admins, and the path from free trial to paid for companies.
  • Production that holds up. Took the platform live on AWS with staging, automatic deploys, alarms and product analytics.
kindredworkplace.com ↗

React · NestJS · Postgres · ElevenLabs voice · Slack & Teams · AWS

Kindred Workplace: measure, then grow

People are measured where they are, coached by an AI that remembers them, and companies see the change.

Where people are
The app web and phone
Slack + Teams coaching in the flow of work
A real conversation uploaded or recorded
Measure
EQ assessment 15 elements of emotional intelligence
Analyze how each person showed up
360 feedback from peers, in rounds
Grow Paige remembers you
Paige, the coach chat and voice
Nudges + to-dos small steps each week
For the company
Team reports where the team stands
Adoption dashboards who's engaged, who's stuck
Kindred Practice: stay present with your clients, Kindred Practice handles the rest
Kindred · formerly Mynd Therapists stay present with clients. The software handles the rest.

Kindred Practice

Live

Top contributor for three years · sessions, AI notes, scheduling, billing

Therapists lose evenings to notes, scheduling and billing spread across a pile of tools. Kindred Practice puts sessions, notes, calendar and payments in one place, and Paige, its AI partner, does the writing.

  • 200+ therapists run their practice on it
  • #1 contributor, 4,400+ commits over 3 years
  • ~40% faster AI answers, with ~30% lower AI costs

What I built

  • Therapy without leaving the app. Video sessions with a waiting room, guest join, consent and recording.
  • Notes that write themselves. Each session is transcribed and drafted into the therapist's own note template, ready to review and sign.
  • Paige, the AI partner. Therapists ask about past sessions in plain language and get answers grounded in their own history.
  • The business side, handled. Calendar sync, a booking widget, reminders, subscriptions, invoices and no-show fees.
  • Less paperwork up front. An AI form builder and e-signed intake forms, plus the move from Mynd to the Kindred Practice brand.
kindredpractice.com ↗

React · Django · Node.js · LiveKit video · Deepgram · Gemini & GPT · Stripe

Kindred Practice: from booking to a signed note

A client books, the session happens on video, and by the time the therapist closes the tab a draft note is waiting.

People
Client books, joins, pays
Therapist runs the practice
The session
Booking + reminders calendar sync, texts and emails
Video room waiting room, consent, recording
Transcript speech to text
AI does the paperwork Paige
Draft note in the therapist's own template
Ask Paige “what did we cover last month?”
The business
Billing subscriptions, invoices, no-show fees
Review + sign notes stay encrypted

what it gives every product

  • A coach that remembers each person
  • Emotion and conversation signals
  • Assessments and scoring
  • One safe door to Claude and GPT
Kindred · shared AI platform Build an AI feature once, and every Kindred product gets it.

Kindred Core

In production

Lead on memory · product integrations

Every Kindred product was rebuilding the same hard AI pieces: memory, safety and model access. Kindred Core does them once, properly, so product teams can ship AI coaching without starting from zero.

  • 1 memory every Kindred coach shares
  • 2 AI providers, with automatic failover
  • 100% of AI calls traced and replayable

What I built

  • Memory for every coach. It remembers what matters, updates facts when life changes instead of piling up contradictions, and forgets a person completely when they ask.
  • Products plugged in. Connected Kindred Workplace's coach and a diet-coaching app, with consent and deletion flowing all the way through.
  • Safe by default. Private details are scrubbed, each product only reaches what it's allowed to, and it fails safely rather than making something up.
  • Cheaper to run. Idle development servers now sleep and wake on the first request instead of running all day.

NestJS · Postgres · pgvector · Claude · Langfuse · Inngest

Kindred Core: the AI every product shares

Products don't each build their own memory and safety. They ask Core, and Core answers the same careful way for all of them.

Products
Kindred Workplace coach memory
Diet-coaching app remembers meals
Partner pilots assessments
Kindred Core
Access each product gets only what it's allowed
Safety private details scrubbed, fails safely
Memory remembers, updates, forgets on request
Models Claude first, OpenAI as backup
Always on
Tracing every AI call can be replayed
Quality checks run before every release

what it does for the team

  • Triages new bugs
  • Plans and ships small fixes
  • Reviews pull requests
  • Drafts support replies
Kindred · engineering automation Bug reports in. Reviewed fixes out.

Agent Ops

In production

Built the agent workflows and knowledge sync

Engineers were losing mornings to triaging alerts and small tickets. Agent Ops hands that work to AI agents that investigate, write a plan, wait for a human to approve it, then open the fix as a pull request.

  • 30+ incidents and tickets fixed by agents each month
  • 4 product teams it works for
  • 6× faster answers from company knowledge

What I built

  • Alert to pull request. Crashes and user reports go straight to an agent that reproduces the problem and proposes a fix.
  • A human stays in charge. Agents research and plan first; nothing is written until someone approves it in Slack.
  • Agents that know the company. A sync that keeps company docs and help articles in the agents' knowledge base.

Inngest · Goose agents · Docker · Claude · Langfuse

Agent Ops: from a bug report to a reviewed fix

Agents do the digging and the first draft; people make the call.

Something breaks
A crash alert
A user report
A small ticket
Agents
Investigate reproduce it, read the code
Company knowledge docs and help articles
Loop
Plan
A human approves in Slack
Fix, as a pull request reviewed, then merged

Enterprise skills platform

Skills assessments for a 25,000-person organization in 14 languages. I run its infrastructure.

Diet-coaching app on Core

Connected a nutrition coaching app to Kindred Core so its coach remembers meals, with consent and deletion built in.

Test automation

End-to-end browser tests and deploy checks for Kindred Practice, so releases stop breaking what already works.

Soonbuilt

My studio. Three live products, all designed, built, hosted and launched by me.

The TallyReel site: sticky-note ideas pinned beside the videos they became
Soonbuilt · Mac, Windows, web, phone One line in, a finished video out. $49 once.

TallyReel

Live

Solo · design, build, launch

Short-form video tools charge every month and meter you in credits. TallyReel makes the whole video on your own computer, script to voice to captions to music, then posts it to your channels at the time you pick, even with your laptop closed.

  • ~90s from one line to a finished Short on a Mac
  • 15 social networks it posts to
  • $0 per video after one $49 purchase

What I built

  • A studio that runs offline. Script, voice, footage, captions and an original music track, all generated on the customer's own machine.
  • You stay in control. Read and edit the script before anything renders, or bring your own.
  • Posting that runs itself. Pin a video to a time slot and it goes out to TikTok, YouTube, Instagram and 12 more, from the web or a phone app.
  • Made for more than English. Hindi and Urdu narration, and desi and Sufi music moods.
tallyreel.com ↗

Python · on-device AI · ffmpeg · Fastify · Next.js · Expo

TallyReel: made on your laptop, posted from the cloud

The expensive part, making the video, runs on the customer's own computer, so there's nothing to meter. Only posting runs on a server.

You
One line “How deep the ocean really is”
On your computer no cloud needed
Script written in ~40s, you can edit it
Voice 13 voices, or your own
Footage + captions word-by-word highlights
Music + render an original track every time
Posting
Your week pin a video to a time slot
Web + phone app approve drafts anywhere
Out
15 networks TikTok, YouTube, Instagram…
The Round Zero homepage: practice the interview before it counts
Soonbuilt · web Practice out loud before the round that counts.

Round Zero

Live

Solo · design, build, launch

Interview prep is question banks and resume puffery; almost nothing lets you rehearse the real conversation. Round Zero tailors your resume honestly, briefs you on the company, then interviews you by voice and shows you exactly where the answer broke.

  • Live voice interviews that adapt and push back
  • 100% of scorecard quotes checked against what you said
  • 0 invented resume claims; it only rewords

What I built

  • A real interviewer, by voice. It adapts to the role, pushes back like a panel, and waits while you think.
  • Feedback you can trust. Every point on the scorecard quotes you, and it's verified against the transcript before you see it.
  • One calm place for the search. Tailored resumes, cited company research, application tracking and progress over time, with a coach that remembers.
joinroundzero.com ↗

React · LiveKit · Deepgram · Claude & GPT · Postgres · Cloudflare

Round Zero: a live interviewer that can't make things up

Prep is grounded in what you paste and what search finds; the interview happens by voice; the feedback is checked against what you said.

Prep
Job + resume
Honest tailoring rewords, never invents
Company brief every claim cited
The interview
You speak live, in the browser
The interviewer adapts, pushes back, waits for you
After
Scorecard each point quotes you, verified
Coach remembers progress across sessions
The Soonbuilt homepage: AI that lives on your computer. Pay once. Keep it.
My studio A one-person studio that ships its own apps.

Soonbuilt

Live

Founder

Soonbuilt makes AI apps that run on the customer's own computer (pay once, keep it) and builds apps for clients. Its free idea search checks every idea against what already exists before anyone spends a cent building it.

  • 3 products launched in 2026
  • ≤3 ideas kept per search, each with proof of a gap

What I built

  • An honest idea search. It researches the web, throws out ideas that already exist, and never shows a link or number it can't source.
  • A plan, not just a list. Pick an idea and get a week-by-week build plan by email.
  • Built to stay cheap. Daily limits keep costs flat; overflow searches run overnight and arrive by email.
soonbuilt.com ↗

Astro · Cloudflare Workers · DeepSeek · Web search

TallyReel shows the script it wrote, with one line edited, before anything renders
You read and edit the script before anything renders.
TallyReel Post: a week board of scheduled videos
Post: pin a video to the week and it goes out on its own.

On the workbench

Prototypes, starters and open-source pieces. Some become products.

Prototype

voice-lab

Voice cloning on an Apple-silicon Mac in English, Hindi and Urdu, no API keys. Urdu is transliterated to Devanagari so the Hindi voice can speak it. The research behind TallyReel's cloned voices.

Python · Chatterbox · pyloudnorm · ffmpeg

Building

ClauseGuard

Contract review for freelancers. Flags risky clauses and missing ones, explains them plainly and suggests redlines. Output held to a strict schema, with precision and recall evals per clause type.

Fastify · React · Zod · evals

Building

DropChat

Chat with your docs from one script tag. Crawl, chunk and embed into pgvector; a Shadow-DOM widget streams cited answers over SSE.

pgvector · SSE · vanilla JS

Open source

mcp-prod

A production template for MCP servers in TypeScript: stdio and Streamable HTTP from one codebase, pluggable storage, tests and CI.

TypeScript · MCP · Postgres · Vitest

Open source

voice-agent-saas

A deploy-anywhere starter for voice agents: LiveKit Agents, Deepgram and Silero turn detection, with transcripts in Postgres. The seed of Round Zero's voice loop.

LiveKit · Deepgram · Postgres

Earlier

Client work and projects from Extendly, Zaytrics and university.