PORTFOLIO & SYSTEM CASE STUDIES

Flagship Products & Architectures

Production-tested software systems engineered by Joshua Ukpai. From behavioral RPG habit engines in Flutter to localized AI intelligence platforms and offline utilities.

Flutter 3.27+ Impeller
Next.js 15 & React 19
Google Genkit & pgvector
Offline-First Drift SQLite
mobile
★ FLAGSHIP
2025–2026

Emerge

Identity-First Habit Engine & Gamified Behavioral Ecosystem

Role: Solo Founder, Systems Architect & Mobile Lead
84%
30-Day Retention
vs 12% industry habit tracker benchmark
15,000+
Habit Completions
Logged with 99.9% offline sync fidelity
<16ms
Frame Render Time
Smooth 120Hz UI via Flutter Impeller
4.8/5
Beta Rating
Across 200+ cohort testers

Emerge fundamentally shifts the habit tracker paradigm from transactional checkboxes to identity-based transformation. Instead of asking "Did you run 5km today?", Emerge reinforces "What vote did you cast today for your identity as an Athlete?". Combining James Clear’s behavioral loops (Cue, Craving, Response, Reward) with RPG mechanics (Archetypes: Athlete, Creator, Scholar, Stoic, Mystic; XP leveling, and biome growth), Emerge makes consistent habit formation engaging and sustainable. Built with Flutter 3.27+ on the Impeller engine for 120Hz micro-interactions, paired with Firebase Gen 2 Cloud Functions and local SQLite/Drift offline caching.

Key Architectural Highlights

  • 5 bespoke RPG archetypes with 4-stage evolving visual avatars driven by daily identity votes.
  • Cognitive implementation intentions engine enforcing "I will [Behavior] at [Time] in [Location]".
  • Digital Locker temptation bundling gating recreational links behind positive habit execution.
  • Habit contracts with automated Cloud Function "Snitch" partner notifications on missed commitments.
  • Offline-first SQLite local persistence preventing network latency or broken streak anxiety.

Technologies & Infrastructure

FlutterDartRiverpodFirebase Gen 2Drift (SQLite)Groq AIRevenueCat
web
★ FLAGSHIP
2025

SocialSense

Social Media Intelligence & Market Sentiment Platform for African SMEs

Role: Solo Creator & Lead Systems Engineer
91.4%
NLP Classification Accuracy
On Nigerian Pidgin and mixed English code-switching
4.2 hrs/wk
Time Saved per SME
In manual complaint triage and competitor monitoring
500k+
Social Posts Ingested
Across 60+ active pilot brands in Lagos and Port-Harcourt
<80ms
P95 API Latency
Across aggregated sentiment time-series endpoints

Emerging market SMEs rely heavily on social media (Instagram DMs, Twitter threads, WhatsApp Business) as their primary sales and customer service channels. However, global social listening platforms (Brandwatch, Sprout Social) cost thousands of dollars per month and fail completely on localized informal communication (Nigerian Pidgin, street slang, code-switched English). SocialSense solves this with an affordable, high-velocity ingestion engine and fine-tuned NLP sentiment classifier that turns chaotic social chatter into actionable business insights, instant churn alerts, and automated weekly WhatsApp summaries.

Key Architectural Highlights

  • Custom fine-tuned NLP model processing Nigerian Pidgin, slang, and code-switched social mentions with 91.4% accuracy.
  • Distributed Celery/Redis crawler handling 50,000+ daily social posts without platform rate limit throttling.
  • Real-time WhatsApp alert dispatch notifying business owners of escalating customer complaints within 60 seconds.
  • Competitor intelligence matrix leveraging pgvector embeddings for semantic clustering of market pricing and friction points.
  • Automated weekly executive digest generating actionable PDF and WhatsApp summary briefs for non-technical founders.

Technologies & Infrastructure

PythonFastAPINext.js 15TypeScriptTailwind CSSPostgreSQLpgvectorRedisCelery
tool
2024–2025

Veridrop

Automated Document & Data Verification Workflow Utility

Role: Creator & Full-Stack Architect
99.8%
Extraction Accuracy
On standardized invoices and KYC forms
<1.2s
Edge Parse Time
Average end-to-end document processing
100%
Zero-Knowledge Privacy
Client-side hashing prevents raw data leakage
10k+
Tested Volume
Processed in stress-test validation suite

Manual document verification for invoices, KYC proofs, and credentials is notoriously slow, error-prone, and fraught with privacy risks. Veridrop provides an instant, developer-ready API and web dashboard for extracting structured data from unstructured documents while guaranteeing zero-knowledge document integrity. Built with Next.js 15, edge functions, and client-side cryptographic hashing.

Key Architectural Highlights

  • Client-side drag-and-drop document hashing (SHA-256) ensuring cryptographic verification before file transfer.
  • High-throughput serverless edge parsing pipeline converting unstructured PDFs into strictly validated JSON schemas.
  • HMAC-SHA256 signed webhook dispatch system with automatic exponential backoff retries.
  • Comprehensive audit log visualizer tracking document verification statuses and cryptographic proofs.

Technologies & Infrastructure

Next.js 15TypeScriptTailwind CSSRadix UINode.js EdgeZodSHA-256 Web Crypto
mobile
2024

LedgerNote

Offline-First Financial Note-Taking & Micro-Accounting Engine

Role: Solo Mobile Architect & Developer
<5ms
Local Query Latency
Sub-millisecond SQLite query execution
100%
Offline Guarantee
Zero cloud servers or network dependency
256-bit
Database Encryption
AES on-device vault security
1,200+
Active Micro-Merchants
Utilizing the offline tool across West Africa

Traditional accounting software is bloated, complex, and requires constant high-speed internet connectivity. LedgerNote allows solo entrepreneurs and market merchants to jot down daily transactions in plain text ("Collected 50k from Emeka for web design", "Paid 12k for generator fuel") and automatically parses them into debit/credit ledger accounts. Built with Flutter, Drift (SQLite), and biometric encryption.

Key Architectural Highlights

  • Natural language heuristic parser converting conversational notes into formal double-entry bookkeeping entries.
  • 100% offline-first architecture with instant sub-millisecond local SQLite data access.
  • Biometrically locked SQLCipher on-device encrypted database ensuring absolute privacy.
  • Multi-currency support (NGN, USD, GBP, EUR, GHS, KES) with offline conversion tables.
  • One-tap clean CSV and PDF statement generation for tax filing and business auditing.

Technologies & Infrastructure

FlutterDartDrift (SQLite)RiverpodLocal BiometricsSQLCipher
READY TO BUILD?

Have a Product Vision You Need Architected?

I partner with early-stage founders, creators, and teams to build production-grade MVPs and mobile apps in 2–4 weeks.