Ken Kariuki
Open to senior / staff roles · remote

Seven live products across trading, media, and automation — each one designed, shipped, and run by one engineer. I care about systems that are transparent and debuggable in production, not just accurate in a notebook.

Founder & Technical Lead · Vancouver, BC · Canada · remote

Live products, operated end-to-end7
Engineers, strategy to on-call1
Core infrastructure<$50/ month
Signals audited to real outcomes640+

Selected work

production, not prototypes

T BOT live

An autonomous, multi-surface algorithmic trading platform. Self-calibrating models behind a 14-layer risk guard chain (position caps, daily-loss limits, per-city guards, slippage), with shadow-outcome backfill so every signal is graded against real settlements rather than idealized fills.

tbot.trade operator dashboard — Portfolio Overview: one card per surface (ST live on Kalshi, FX, CRYPTO, LT, STK), health, and the labeled revenue projection
tbot.trade — the operator dashboard, one card per surface
tbot.trade ST surface page — bankroll mark-to-market, banked P&L, max drawdown, equity curve, positions, performance and the Kalshi settlement resolver
ST surface — bankroll, P&L, drawdown, resolver
tbot.trade/demo — the public product preview: algorithmic trading signals delivered to your phone, 15% on realized profit only, $0 custody
tbot.trade/demo — public product preview
PythonFastAPILLM orchestrationquant / dataDigitalOcean

JobScout live demo

A self-running job-search pipeline: discover → sift → score → apply. Sweeps ~1,400 postings a day, scores fit honestly with an LLM, and hands the judgment calls to a Telegram control surface. The public demo scores today's real sweep live — gates, reasoning, and costs on screen.

JobScout on Android — the day's real sweep streaming the eligibility gates, 14 postings cleared, scoring live
Native Android (Kotlin · Jetpack Compose), real device
JobScout on iOS — the candidate chooser with the day's real feed loaded, captured by the CI simulator step
Native iOS (Swift · SwiftUI), simulator via CI
PythonClaude APICloudflare Workers + D1Kotlin · ComposeSwift · SwiftUITelegram Bot API

T BOT Web Studio client sites

Client websites designed in Chat and shipped by Code on Cloudflare Pages — fixed-price builds plus a recurring care plan. Latest: Sacred Earth, a Uganda travel brand's field-guide site — a scroll-driven altitude sweep over a canvas point-cloud of the country, fifteen self-contained pages generated from one builder, self-hosted fonts, zero runtime network calls, legal copy ported verbatim, every ship link-checked and console-clean before it goes live.

Sacred Earth — a digital field guide to sacred ground: the Descent homepage, a canvas point-cloud of Uganda swept by altitude
sacred-earth.pages.dev — the Descent
Cloudflare Pagescanvasstatic-firstdesign → build handoffclient care

Unified Ops Dashboard edge-native

One signed audit table for an entire product portfolio. An HMAC-signed event-ingest Worker writes every product's events into a shared Cloudflare D1 store, feeding an operator dashboard and a subscriber portal, end-to-end verified across platforms. Built so a value can always be traced to a signed source, never recomputed.

ops.tbot.trade — the operator dashboard behind Cloudflare Access: fleet health pills, flags, scheduler queue, one tile per product, then per-project System health cards
ops.tbot.trade — fleet overview (behind Cloudflare Access)
ops.tbot.trade — Money & funnels: one revenue card per surface, deposits never counted as P&L, one owner per fact
Money & funnels — one owner per fact
Cloudflare WorkersD1PagesHMACTypeScript
ops.tbot.trade

OddsEdge live

An automated sports-intelligence subscription. Deterministic de-vig fair pricing from sharp-consensus odds (no LLM guesswork on the numbers), a daily brief that self-publishes across channels, and a closing-line-value rail that keeps the edge honest.

oddsedge.win — an AI analyst for sports value betting: best price vs consensus-implied fair, the edge quantified, no invented picks
oddsedge.win — the daily brief
PythonCloudflare PagesWhop billingautomation

Content Studio autonomous

An end-to-end AI video production line: screenplay → keyframes → motion → proofread gates → publish, running unattended across multiple channels, with a validator that refuses a render before it wastes generation credits.

Zamani on YouTube — Africa, before you were told: a new film every Sunday, 61 videos, produced end-to-end by the pipeline
Zamani (@ZamaniHistory) — a film every Sunday
The Fortune Files on YouTube — the full arc of money and power from court records and SEC filings, a new file every Thursday
The Fortune Files (@fortunefiles_HQ) — every Thursday
Nodegenerative videopipeline designQA gates
private

Stack

Languages
PythonTypeScriptJavaScriptSQLBash
Edge / Cloud
Cloudflare WorkersPagesD1R2KVDigitalOceanFastAPInginx
AI / Data
LLM orchestration (Claude)agentic pipelinescalibration & backtestingquant / signal
Product
Capacitor (iOS/Android)StripeWhopTelegram botsOTA delivery
Security
ICS / SCADA / HMI hardeningnetwork segmentationCEHSecurity+Network+

How I work

Instrument everything

If it runs in production, it emits a signed signal. Silent success is invisible: I build the sensor before I trust the system.

Honest by construction

Numbers reconcile across surfaces or they don't ship. Shadow-grade against real outcomes; measure before believing an edge.

Cost is a constraint, not an afterthought

A seven-product portfolio on under $50/month of core infrastructure. Tight budgets force clean architecture and real discipline.