Jayenne Montana

• Truth, enforced not trusted. The social graph acts as a live validation engine for the dating graph. A decoupled service layer feeds behavioural signals to AWS Bedrock for persona inference and stores versioned snapshots, so when a profile drifts from a member's real behaviour over a trailing month, a background pipeline flags it.
• A reputation cascade modelled on PageRank. Standing travels up each invitation lineage, so inviting a bad actor costs the inviter 15% of their own standing. The network self-moderates at the point of entry, because vouching for someone means something again.
• A purpose-agnostic matching engine. Strategy patterns and exponential decay curves instead of brittle binary filters, with bidirectional strictness so a match never scores higher than the less keen party. That quietly kills false positives.
• Values that outrank the algorithm. A sovereign disinterests rule means a member's explicit boundaries can never be overridden by whatever would keep them scrolling. The feed will not radicalise people for engagement, by design.
• Security treated as mathematics. Blind-indexed, CipherSweet-encrypted personal data, liveness checks, and location fuzzing that deliberately never re-fuzzes, because averaging samples is exactly how an attacker triangulates a real position.
• Infrastructure that respects the bill. Media moderation runs a cost-aware waterfall on AWS Rekognition and only escalates to a second provider when explicit-content ratios spike. The daily prompt schedules no per-user database rows at all, computing its timing from a hash to avoid a thundering herd.

Live and running. Private repository access available to seriously interested parties. [ADD: live URL]

• A national-scale geospatial engine. Built and maintained Laravel and Vue/React products delivering sophisticated backend analytics through SaaS platforms and robust APIs over a trillion-record PostGIS data store.
• Marketplace expansion. Spearheaded a standalone Laravel/React application for the ReapitCRM marketplace, embedding compliance tools directly into the workflows of letting agents and property managers and significantly extending Kamma's client reach.
• Partner SDKs. Designed RESTful APIs and the associated SDKs that let third-party partners integrate Kamma's intelligence into their own systems, solidifying its position as the industry's primary data provider.
• Retrofit and energy efficiency. Built a tool that translates abstract EPC data into an actionable roadmap, helping homeowners reach specific EPC bands or budget goals and generating optimised retrofit packages, in direct support of the UK's move to carbon neutrality.
• Clean code as policy. Maintained clean, tested code while evolving the core platforms from legacy versions to modern, performant architectures.

My work supported Kamma's recognition as "Best ESG Data Provider" and its Green Homes Finance Accelerator funding for retrofit optimisation.

• An autonomous data discovery and enrichment engine. Resource-grouped ingestion queues prioritised by data richness and staleness, with silent-zero detection, stall-cluster reclaim and live dashboard streaming.
• A disciplined pipeline. Collection, sanitisation, normalisation and a three-stage validation cascade, using a first-writer-wins provenance pattern to eliminate duplicates at scale.
• A scoring engine. It reads Companies House name-change histories and multi-office footprints to infer appetite, ability to buy and succession risk, matching practices closely enough to make post-deal integration smoother and cheaper.
• A deal state machine. Governed, conditional progression with anonymous teasers, e-signable NDAs, CipherSweet-encrypted financials and secure S3 vaults behind signed URLs.
• A high-deliverability outreach engine. Built on the principle that volume is vanity. Content variants and send windows are both treated as arms in a Thompson-sampling model, tuned by daily feedback on reply warmth and bookings, skipping weekends and bank holidays. It pairs live SpamAssassin scoring with an asynchronous AI pipeline that scores reply intent, funnels warm leads to the right people and unsubscribes hostile responses automatically.

Built under contract for a venture in the UK accountancy sector and currently in its final stages of development. I can discuss the engineering and results in depth, and arrange access, with serious enquiries.