← Pemystack

Case Studies & Projects

Work built around quality at every layer.

Consulting engagements, live AI platforms, and automation tools, all grounded in the same principle: measurable quality outcomes over process theatre.

Full-stack products built end-to-end

Production SaaS platforms covering architecture, data layer, AI integration, payments, and deployment.

Live LegalTech · SaaS · Kenya

CaseLens

Full-stack development · AI-powered legal intelligence · legal.pemystack.com

CaseLens hero

A production LegalTech SaaS for Kenyan advocates. Supports three practice modes (prosecution, defence, and civil), each generating AI-drafted legal documents anchored in Kenyan statute: ODPP format briefs, bail applications, no-case submissions, plaints, and written submissions.

Full-stack Next.js App Router, PostgreSQL via Neon + Drizzle ORM, AES-256-GCM encryption at rest, M-Pesa and IntaSend payments, Tesseract.js OCR, JWT auth with email verification, Playwright and Vitest testing, and Vercel Cron for subscription lifecycle and demo resets.

3 practice modes (prosecution, defence, civil)
3 AI providers (Claude, Groq, OpenAI)
M-Pesa live payment integration
Live deployed at legal.pemystack.com
Live Audit · SaaS · Kenya

Vigezo Audit

Full-stack development · ISA-compliant audit document generation · audit.pemystack.com

Vigezo hero

A SaaS platform built for Kenyan audit firms. Auditors upload trial balance files from QuickBooks, Sage, or Xero and AI generates complete, ISA-compliant documentation: working papers across 15 audit areas, audit reports with automatic opinion determination, engagement letters, and management letters.

Multi-provider AI (Claude, GPT-4o, Groq — configurable per firm), M-Pesa and IntaSend subscription payments, client portal for secure document sharing, team role workflows, and Kenya-specific compliance logic (ICPAK, KRA, ITAX, Companies Act 2015).

15 audit areas covered end-to-end
4 ISA document types generated by AI
3 AI providers (Claude, GPT-4o, Groq)
M-Pesa live payment integration

Conversational AI platforms for the Kenyan market

MCP-backed AI assistants that let operators query live business data in plain English or Swahili.

Live Logistics · AI · Kenya

Msafara Logistics

AI-powered logistics platform for the Kenyan freight and supply chain ecosystem. Operators query shipment status, fleet compliance, KRA customs clearance, M-Pesa payments, and warehouse inventory in plain English or Swahili, against live PostgreSQL data.

15 FastMCP tools cover every layer of operations, from Mombasa Port IDF numbers and NTSA compliance to Nairobi ICD and SGR rail. File upload lets analysts interrogate their own Excel and PDF reports inline.

FastMCP Groq AI FastAPI React Neon
View live →
Live FinTech · AI Assistant · Kenya

Akili Finance

AI-powered assistant for Kenyan SACCOs. Staff query members, loans, and financials in English or Swahili. PAR analysis, risk scoring, SASRA returns, and IFRS 9 ECL reports, all from live data via 14 MCP tools with five-tier RBAC.

React + Vite frontend, FastAPI streaming backend, Groq (Llama 4) inference, Neon PostgreSQL. File upload lets analysts query their own Excel, CSV, and PDF reports inline.

FastMCP Groq (Llama 4) FastAPI React + Vite Neon
View live →
Live EdTech · AI Platform · Kenya

Skooli Assistant

AI management platform for Kenyan schools. WhatsApp-driven attendance roll call for teachers, parent self-service portal with M-Pesa fee payments, AI analytics chat against live school data, and a principal admin panel, all on phones staff already use.

FastMCP tools, FastAPI + SSE streaming, React + Vite frontend, Meta WhatsApp Cloud API for notifications, Safaricom Daraja for M-Pesa payments, APScheduler for automated pings and weekly reports.

FastMCP Groq / Claude FastAPI WhatsApp API M-Pesa
View live →

QA strategy, automation, and AI-assisted testing

Operating model transformation, risk-based automation design, and intelligent CI/CD tooling.

Delivered Strategy · Leadership · Quality Engineering

Quality Assistance Model

Lead Software QA Engineer · Engineering-wide transformation · Trellis

Replaced a siloed QA gatekeeping model with a Quality Assistance approach that shifted ownership into the engineering team, rebalanced automation around risk, and made release confidence part of the delivery system, not a late-stage approval ritual. The change touched how teams planned, how PRs were reviewed, how incidents were analysed, and what "done" actually meant.

80% fewer post-release defects
70% faster test execution
20% faster time-to-market
40% year-on-year efficiency gain
Active Android · CI · AI-Assisted

Android PR Acceptance Testing Agent

Claude-powered GitHub Actions agent that selects which of 60 instrumented Android tests to run per PR using diff analysis, runs them on a managed emulator across up to 4 shards, then diagnoses failures and posts coverage gap recommendations directly to the PR comment.

Claude API GitHub Actions Android Emulator Node.js
Read more →
Live AI Workflow · Code Review

LLM PR Regression Checker

Autonomous pull request review agent that compares code diffs against a vector-backed library of historical bugs, then posts evidence-based regression feedback directly into GitHub review flows.

Python OpenAI GitHub API Sentry
Read more →

Work Together

Bring Pemystack into your next project.

Whether you need a quality strategy review, an automation build, or an AI-assisted workflow. Let's talk about what measurable looks like for your team.

Start a conversation
← Back to Pemystack