Our client, a Swiss EV charging infrastructure provider operating more than 5,000 chargers across Switzerland, was processing over a million charging sessions a month, but their data was pull-only. Fleet managers, site hosts, and drivers had to log into a web dashboard every single day just to check consumption, cost, or whether a charger had gone offline overnight. Nothing reached the customer unless they went and looked. We built a remote MCP server in TypeScript on Node.js, hosted on AWS ECS inside the client's VPC, that connects Claude and ChatGPT directly to the client's existing REST API. Authentication reuses the client's existing AWS Cognito user pool, extended to support the OAuth 2.1 authorization code flow with PKCE that MCP requires, so customers connect using the same dashboard credentials they already have, no new account needed. Each MCP tool is a thin wrapper around an existing endpoint, with no duplicated business logic. Phase 1 shipped ten read-only tools (consumption, cost, charger status, alerts, session history), and Phase 2 added session start/stop and alert-threshold controls, each behind an explicit human confirmation step. Results: 5,000+ chargers and 1M+ monthly charging sessions now queryable in natural language, by the same 10,000+ monthly active users, with no new backend and no new identity store. Daily dashboard logins have been replaced by recurring prompts for consumption, cost, and offline alerts. The client was the first in its competitive set to ship an AI-assistant integration, delivered by a team of just two engineers.
MTechZilla
RightFirms reviewed this company’s profile, work, reviews, clientele, and public trust signals.Software Development Company for Product & Engineering Teams
MTechZilla Portfolios
Our client, a Swiss EV charging startup, needed to process utility invoices for installations spread across Switzerland, arriving in four languages, in layouts ranging from 3 to 10 pages, from hundreds of different local utilities. Since these figures fed directly into billing, accuracy couldn't be compromised, but the only existing process was an admin manually reading and typing every figure by hand. We built an AI-assisted invoice capture system combining deterministic parsing (reading Switzerland's QR-bill standard directly from the QR code for the gross amount) with GPT-5.4 for fields with no machine-readable source. We also inverted the old workflow: instead of an admin picking an installation first, the system now infers which installation an invoice belongs to directly from the document, and resolves it live on screen for the admin to confirm. Critically, the AI never submits data on its own, every extraction is shown beside the original PDF for human confirmation before anything is saved. Results: Handles 4 languages (German, French, Italian, English), parses 3-10 page invoices in a single pass, and maintains 100% human verification on every financial figure before storage. The system later expanded to suggest per-kWh pricing markups based on confirmed invoice data, comparing what utilities charged against what was already collected from drivers.
Our client, a US-based hospitality startup working with 20,000+ hotels nationwide, was drowning in document chaos. Every hotel partner sent invoices, folios, receipts, and statements in its own format, and traditional OCR failed so often that a large reconciliation team had to correct the output by hand. We built an AI-powered document processing pipeline using OpenAI's GPT models with function calling, replacing brittle OCR templates with a single system that reads any document format. Every page is rendered to a high-resolution image and sliced into overlapping segments so nothing is lost at the cut lines, then read against a strict extraction schema. Built on AWS (Lambda, SQS, S3, DynamoDB), the pipeline scores every extraction for confidence, automatically routing uncertain documents to human review, whose corrections feed back into the system as few-shot examples to keep improving accuracy. The pipeline launched on GPT-4.1, and upgrading to GPT-5.4 was a configuration change, not a rebuild, since extraction is fully schema-driven. Results: 100,000+ documents processed monthly, extraction accuracy improved from 80% in the pilot month to 95% at full scale, supporting 20,000+ different hotel document formats. Delivered in 4 months by a team of 4 AI engineers.
Our client, a US-based social services technology platform, helps agencies place vulnerable people into emergency shelter, but that only works if they know real-time bed availability. Since American shelters publish no API or data feed, that information only exists if someone calls and asks. The client was paying a call center to manually dial shelters city by city, every day, an operational cost that grew with every new city covered. We built an AI voice calling system using Retell AI and AWS Lambda, with calls dispatched via Amazon EventBridge Scheduler on timezone-aware schedules, so each shelter is reached during local working hours. The system extracts structured data (bed counts, restrictions, intake notes) directly from real call transcripts, not model guesswork, validates it against known shelter capacity, and automatically retries unanswered calls or voicemails. Human spot checks on sampled transcripts keep quality honest as coverage scales. Results: Hundreds of shelters called daily by AI voice agents, scaled nationally from a single-state pilot in Texas, delivering daily bed-availability updates without a single human dialing a phone.
Sintio set out to compete in the Swiss EV charging market with almost no software to its name. They needed an entire technology ecosystem built from scratch, and a team that understood EV charging deeply enough to do it right. We built all of it: a CPO management portal, a public charging web app, native iOS and Android apps, and the marketing site, delivered through one unified engineering engagement. Starting from discovery and Figma-based UX, we shipped a functional MVP in 8 weeks, so Sintio could onboard its first customers while the rest of the platform was still being built. The technical core is a universal OCPP gateway. Most charging platforms need bespoke integration work for each hardware brand; ours uses manufacturer-specific adapters to support 30+ makers out of the box, including Zaptec, ABB, Hager, Mennekes, and Easee. Sintio can onboard virtually any charger without custom engineering. For drivers, we removed the biggest barrier to public charging. Scan a QR code, pick a payment method, start charging — no app download, no account. The flow supports TWINT, Visa, Mastercard, Amex, Apple Pay, and Google Pay with sub-second response at peak load. It runs on a Node.js and AWS backend, with React, Next.js, and React Native across the front ends and Stripe handling billing. Today the platform manages 5,000+ charging stations across Switzerland, processes 1M+ sessions a month, and serves 10,000+ active monthly users. Sintio used that early traction to secure funding and expand across residential, commercial, and public charging. The 8-week MVP has become a multi-year partnership. "The commitment of their developers to our project has been outstanding." — Philipp Bruhin, Head of Software Development, Sintio AG
When a domestic violence survivor needs to leave tonight, a hotel's ID and credit card policy shouldn't be what stands in the way. That was ReloShare's problem to solve: caseworkers at social service agencies were spending hours on the phone trying to place clients in emergency housing, often blocked by requirements those clients couldn't safely meet. We joined as an embedded extension of ReloShare's engineering team when they were a two-person startup, owning UI/UX and custom development. Our core deliverable was Safe Stays, a platform that lets agencies book hotel rooms under client aliases with no ID, credit card, or deposit required. The hard part was identity. We built a backend system that keeps a client's real identity fully protected while satisfying each hotel's compliance and check-in requirements. Around it we shipped consolidated monthly billing across hundreds of agencies, an instant availability search engine, a fixed-fee protection plan that ends surprise charges, and 24/7 check-in confirmation workflows. It runs on a serverless AWS stack (Lambda, API Gateway, DynamoDB) with a React and Next.js front end, delivered through agile bi-weekly sprints. The outcomes are measurable. Safe Stays has facilitated 500,000+ safe nights, onboarded 700+ agencies nationwide, and built a network of 15,000+ hotels including Choice Hotels, Wyndham, and Extended Stay America. Work that once took hours now takes seconds, freeing caseworkers for the people they serve. Across our ongoing partnership, ReloShare has grown from two people to a 40-person company. "They cared about the work we were doing, and we considered them part of our team." — John Moats, CTO, ReloShare
For a furnished-housing operator, the website is the storefront and the lead engine. Pinnacle Furnished Suites came to us to rebuild theirs from the ground up, into a platform worthy of 50+ luxury buildings across 11 Chicago neighborhoods and Nashville, including towers like AMLI River North, Aqua, and Wolf Point West. We delivered the full rebuild in two months: discovery into their sales workflow and tenant journey, Figma design, then a fast, SEO-optimized build. Corporate travelers and relocating professionals can search by neighborhood, building, unit size, and guest count, while an interactive map shows exactly where each property sits relative to business districts, transit, and landmarks. Two decisions did the heavy lifting on business impact. A streamlined quote request form captures move-in dates, unit size, and guest count, then notifies the sales team instantly, cutting friction out of the lead pipeline so prospects reach a human faster. A tenant portal gives renters self-serve access to their account, payments, and services, reducing the administrative back-and-forth that used to eat staff time. We built it on Next.js, Supabase, TailwindCSS, and Vercel, with Google Maps for geographic search. The Supabase-backed CMS lets Pinnacle's team add new cities, buildings, and units on their own, with no developer involvement, so the platform scales with the business instead of waiting on us. "They are super responsive, fast, and just a great organization. I highly recommend them for projects, scaling your development teams, or as your entire team." — Matt Singley, CEO, Pinnacle Furnished Suites
Our client, a venture-backed, seed-stage US PropTech startup, set out to bring pricing transparency to the rental market. Landlords, tenants, property managers, real estate investors, and insurance adjusters all lacked access to reliable, unbiased rental valuation data, despite significant price variation even within the same neighborhood. We built a unified API integration layer that aggregates and normalizes data from multiple premium property valuation providers, cross-referencing recent transactions and active listings. Our algorithm weights property size, location, amenities, and market trends to generate instant, detailed rental estimates covering all 50 states. Built on React, Next.js, and Supabase/PostgreSQL, delivered as an MVP in 6 weeks. Results: 10M+ properties covered, 10,000+ monthly reports, landlords reduced vacancy rates by 35%, tenants saved an average of $2,400 annually, achieved product-market fit in 4 months with a 4.7/5 user satisfaction rating, and helped the client secure their seed funding round.
Our client, ReloShare (a Chicago-based social services technology company), identified a critical gap beyond their hotel booking platform: agencies had no way to see real-time shelter bed availability, forcing case managers to call shelters one by one, a process that could take hours during a genuine crisis. We built The Grove, a two-sided, event-driven platform where shelter providers update live bed counts in seconds, while agency case managers search the entire network at once with flexible filtering (age, gender, family configuration, accessibility, sobriety requirements). De-identified referral workflows protect client identity throughout, critical for domestic violence survivors and trafficking victims. Built on React, Next.js, Node.js, and AWS (Lambda, API Gateway, DynamoDB). Results: 20,000+ shelter beds managed, 700+ agencies connected, 95% reduction in placement time. Real client feedback (5.0/5.0 on Clutch, John Moats, CTO): "We grew from a 2-person startup to a 40-person team with the help of MTechZilla."
Our client, Strompreise Schweiz, needed to solve a real transparency gap in the Swiss energy market. Switzerland has hundreds of regional distribution system operators (DSOs), each setting their own complex pricing across energy tariffs, capacity pricing, grid usage, and meter fees. The only public reference, a government portal (ElCom), published only averaged values, too blunt for modern smart energy systems like EVs, heat pumps, and solar inverters that need precise, real-time pricing to optimize consumption. We built a Next.js frontend for tariff browsing and Node.js backend services on AWS with DynamoDB, delivering machine-readable RESTful APIs fully compliant with Switzerland's Electricity Supply Ordinance (StromVV) Article 7b. The platform covers 600+ energy providers with granular, itemized pricing data smart devices can query directly. Results: 600+ Swiss energy providers covered, full machine-readable API access, delivered as a production-ready MVP in just 1 month. Client feedback (5.0/5.0): "MTechZilla increased our development velocity and cut backlog resolution time by over 30%... it truly felt like they were part of our own team."