Senior Full-Stack Engineer | Mobile Attribution Platform (iOS SDK + Backend) | 1-Month MVP Contract

Modern MMP's charges $2,000+/month and still cannot tell you which ad actually drove a paying subscriber. Their data is broken by Apple's ATT privacy framework, their SKAdNetwork implementation is a black box, and their influencer attribution is essentially guesswork.About SiraWe are building Sira: a post-ATT attribution platform for iOS apps that is more accurate, more transparent, and built specifically for the era where IDFA is dead.We are looking for one exceptional engineer to build the core attribution engine from the ground up and make it genuinely better than what exists today.The Problem We Are SolvingAfter Apple's ATT, every major MMP is struggling with the same core issue: they cannot deterministically match ad clicks to installs anymore. Their solution has been to bolt SKAdNetwork on top of their old architecture and call it a day. The result is coarse, delayed, and often wrong attribution data.The right solution is a new attribution framework built from scratch — one that treats SKAdNetwork as the primary signal, layers probabilistic fingerprint matching on top, and uses AI to fill the gaps where neither can reach. The goal is not to replicate AppsFlyer. It is to make them obsolete for iOS-first apps.What You Will BuildiOS SDK (TrackKit — Swift Package Manager)it(apiKey:) on first app open — registers the device, captures anonymous fingerprint (IP, device model, OS version, screen resolution, timezone, locale), stores UUID in KeychainSKAdNetwork.registerAppForAdNetworkAttribution() on installStoreKit 2 transaction observer — fires purchase events to backend in real timeDeferred deep link capture for influencer campaign attributionDrop-in simple: one line to initialize, zero configuration requiredAttribution Engine (Backend — Node.js/TypeScript)SKAdNetwork postback ingestion with Apple ECDSA signature verificationProbabilistic fingerprint matching: when a user installs the app, match their fingerprint against the click log within a configurable lookback window. Score the match probabilistically, not binary.AI-assisted attribution layer: when SKAdNetwork and fingerprint matching both fail or conflict, use a lightweight model trained on historical install patterns to infer the most likely source. No MMP does this today.Deduplication: one install is never attributed to two sourcesConfidence scoring: every attribution result gets a confidence score (0–1) so customers know how reliable the data isAd Platform IntegrationsMeta Ads API — OAuth connect, pull campaign spend and impressions dailyGoogle UAC API — OAuth connect, pull campaign spend and installs dailyApple App Store Connect API — pull revenue, proceeds, and unit salesDashboardReact-based analytics dashboard showing ROAS, CPI, LTV, net profit, and channel breakdown — all wired to real attribution dataYou Must HaveSwift — you have shipped an iOS SDK or Swift Package before, not just an appNode.js + TypeScript — backend API development, event pipelines, database designSKAdNetwork — you understand the postback flow, privacy thresholds, coarse/fine conversion values, and the 24–48 hour delay. If you have to Google what a postback sequence index is, this is not the right role.Probabilistic matching — you have built or worked on a fingerprint or device-graph matching systemOAuth 2.0 — you have integrated at least one major ad platform API (Meta, Google, or TikTok)Cryptography basics — JWT, ECDSA signature verification, webhook securityStrong Advantage If You HavePrior work at Adjust, AppsFlyer, Branch, Singular, Kochava, or a mobile growth agencyExperience with ML-assisted attribution or probabilistic inference systemsFamiliarity with Apple Server Notifications and StoreKit 2A strong opinion on why current MMPs get post-ATT attribution wrongStructure1-month contract,$3,500 fixedFully remote, async-first.IST or EST timezone preferredLong-term potential if the product gains traction, but no commitment beyond the initial month.How to ApplyAttach ResumeAlongside in the resume, add a Loom somewhere (you can add it in a link inside your resume) answering the following question:/"SKAdNetwork 4.0 gives you a coarse conversion value of 'medium' and a 48-hour delayed postback. The ad network claims credit for the install. Your fingerprint match gives you 73% confidence it was organic. How do you attribute this install and what do you show the customer?/"

Apply Now