Most engagements under NDA — details available on request.
SYS — PROJECT CANVAS
Where AI stops and humans decide
We don't delegate the decisions that matter.
AI
AI handles
Scaffolding, boilerplate, test generation, UI iteration, code review suggestions
Humans decide
Architecture, data model, security, UX choices, final QA, deployment sign-off
AI
AI handles
Documentation, API spec generation, regression test suites, changelog drafts
Humans decide
Product direction, client communication, risk assessment, IP strategy
IP never leaves your project's context. No third-party AI training on your code. See our full approach →
Tier 1 — AI-Native Full Ownership
Built end-to-end with AI-native workflows
These are the projects we owned completely — from discovery to deployment — using AI coding agents throughout. Architecture, data model, and final QA stayed human. Everything else moved faster. The "What AI handled / What humans decided" split below is not marketing language — it's the actual decision log. How AI changes our workflow →
Flagship Case Study — Full Ownership · AI-Native
Interactive Experience · Logistics AIJan 2026
Enmovil AI — Interactive Product Experience
Enmovil Solutions Pvt. Ltd. — India
The problem before us
Enmovil — a supply chain and logistics AI platform — needed to demonstrate the depth of their product suite in a way that a static website simply couldn't. A sales deck wasn't enough. They needed something prospects could feel, not just read. The challenge: convey six distinct product chapters (Supply Chain Planning → Logistics Execution) with real demo videos, without losing visitors to a passive marketing page.
What AI handled
Full UX design in Figma — Claude and ChatGPT drove every wireframe iteration, zero traditional design agency involved
Next.js + TypeScript scaffolding, component architecture, and CSS module structure
Users scroll (or swipe on mobile) to scrub through a looping video timeline. At key timestamps, interactive hotspot callouts appear over the video — clicking one slides in a panel with a Wistia product demo. The scroll engine uses velocity-based physics with separate tuning for mouse wheel, trackpad, and touch, so the experience feels intentional on every device.
Every hotspot interaction — open, dwell time, close — is tracked via a custom analytics schema, giving Enmovil's team a live view of exactly which product demos prospects are most engaged with.
When a product gets recalled — by the government or the retailer — the people who already bought it have no idea. There's no reliable channel from "recall issued" to "consumer notified." People find out by accident, or not at all. The client wanted to close that gap with a mobile-only app that sits quietly in a user's pocket and screams when it matters.
What AI handled
OCR receipt scanning scaffolding — 2 days vs ~1.5 wk manually
Test suite generation for the ML barcode + receipt pipeline
API integration boilerplate for 3 government recall databases
UI iteration — same-day turnaround per feedback cycle
What humans decided
ML model selection — accuracy threshold before a notification fires
Push notification architecture and latency budget (false negatives vs. false alarms)
Data privacy model — what gets stored, what gets hashed, what never leaves the device
Final QA — every edge case (torn receipts, partial barcodes, multi-item scans) reviewed by a senior engineer
98%
OCR scan accuracy
80%
Manual recall-check work eliminated
<2 min
Recall alert after notice posted
How it works
User scans a receipt or barcode with their phone camera. The app extracts product identifiers via OCR and matches them against a continuously-synced index of active government and retailer recall notices. If a match is found — against the same batch, packaging series, or UPC — a push notification fires immediately. No manual checking, no browsing recall websites.
The pipeline runs entirely on the device for the scan step. Matching happens server-side against the recall index, which is polled from government APIs on a sub-5-minute refresh cycle. Mobile only — iOS and Android.
React NativeiOS / AndroidPython (ML)OCR PipelineNode.jsPostgreSQLAWSPush NotificationsGov API integrations
Flagship Case Study — Co-Founded, Full Ownership
Web (Desktop) + Mobile App · EdTech2024–2025
EduHire — Teacher Recruitment Platform
Practills Learning, India — Co-founded with Deepika (CEO)
The problem before us
Academic hiring is broken in a specific, fixable way. Schools post on generic job boards and wade through unqualified applicants. Teachers send CVs into voids. Neither side has a purpose-built tool — every existing platform is repurposed from corporate hiring and completely wrong for the rhythms of academic recruitment: semester timelines, subject specialisations, institutional trust. We built the dedicated layer that was missing.
What AI handled
Full platform scaffolding — functional in first 10 days
Teacher–school matching algorithm iteration and scoring logic
WebRTC interview integration boilerplate and test coverage
Iteration cost reduced ~40% vs a traditional agency approach
What humans decided
Matching criteria and weighting — a product decision grounded in how schools actually shortlist
Dual UX — desktop-first for school admins, mobile-first for teachers on the go
Keeping scope tight — no unnecessary features, purpose-built for teacher recruitment only
Schools create a position with subject, grade level, location, and requirements. Teachers build a profile with qualifications, experience, and availability. The platform surfaces ranked matches on both sides — schools see suitable candidates, teachers see relevant openings. Applications, shortlisting, and interview scheduling all happen inside the platform.
Deliberately scoped to academic recruitment only — no freelance gigs, no corporate jobs. School admins work on desktop; teachers apply and track progress on the mobile app. No bloat, no generic job board features that don't apply.
Want to understand the full AI workflow behind this project? Read the breakdown →
Tier 2 — Traditional Builds (2018–2024)
Senior engineering before AI tooling existed
These projects were built by senior engineers the traditional way — no AI coding agents, no scaffolding shortcuts. They still run. They still hold up. They're also the reason our current AI-accelerated work is trustworthy: when you've hand-written the same code both ways, you know when to trust the machine and when to override it.
Traditional — senior engineering, no AI tooling
Mobile App · Energy2020 - Present
Eflex Clean Energy System
Traditional Build
Andrew Winter, CEO — Eflex UK
Intelligent residential and commercial energy management — controls solar, battery, and grid power from a single app. Built 0→1, still running, still improving.
Complete restaurant management — web POS, mobile ordering, inventory tracking, and analytics dashboard. Built for an Australian chain with multiple outlets and real-time sync.
ASP.NET Core.NET MAUIAzureSQL ServerStripe
40% faster order processing · 99.5% inventory accuracy
Full ownership — POS, mobile, analytics
Web App · Logistics2023
Harbour Vessel Management System
Traditional Build
Christopher — Logistics Co.
Full digital port management: freight, load optimisation, document clearances, real-time vessel tracking. Replaced paper-heavy processes for a busy harbour operation.
End-to-end e-commerce solution for one of South India's most recognised clothing retailers — website, mobile app, payment gateway, inventory, and SEO from scratch.
Social media-inspired learning platform built during COVID-19 — voice chat, engagement tools, an interface gentle enough for first-time internet users. Tested with 500+ students across Tamil Nadu.
AngularASP.NETKafkaElasticsearchAWS
500+ students · 85% engagement rate · 4.5★ satisfaction
Internal — founder-owned and operated
Web App · Analytics2021
1vYou — Data Visualisation Platform
Samerfam, Dallas USA
Upload any dataset, generate interactive charts and live maps in seconds. Built for researchers and analysts who need answers fast without engineering support.
Full ownership — product design through deployment
SEO & Growth · Media2021
Voice of India — News Aggregator
Internal Project
News aggregation platform — automated curation, clean editorial UI, and a full SEO strategy that drove 500k monthly page views within six months of launch.
Leading US Pediatric Health System (via Element Blue), USA
Modernized the patient portal for one of America's largest pediatric health systems. HIPAA-compliant, Sitecore-based, fully accessible. 100k+ patients onboarded at launch.