Proof work should change what a team decides next.

Medical carts needed safe hot-swappable power across multiple voltage outputs, while batteries and charging stations also needed remote monitoring.

Advised on product architecture across smart BMS, protection, hot-swap, 3–70 V and USB-C outputs, charging stations, and Wi-Fi/cloud monitoring; supported the roadmap and engineering hiring.

Connected the requirements for 36 V, 680 Wh batteries and one-, two-, and four-battery charging stations in one implementable hardware/software roadmap.

Electricity, heat, and water readings had to be collected, stored, and analyzed without frequent field maintenance.

Designed the system across low-power sensors, Lumiot base stations, GSM-connected gateways, cloud device management, data collection, and visualization; LoRa support was planned.

Specified up to 2,000 devices within a 4 km radius per base station and a sensor design rated for 14 years on a single AAA battery.

RedSwan needed to put real estate in front of its investors, first as a mobile app and then as a web platform built around tokenized offerings.

Shipped a privately distributed iOS app in 2015, then designed the platform from zero in 2018: investor cabinet, property listing and every flow between them. Led it as technical product manager, hiring and directing two developers inside a four-person team.

A working platform for RedSwan's marketplace and a collaboration that ran until around 2023. Payments were deliberately left out at that stage so the flows could be tested first.

Prospects had to trust the interaction and sales teams had to find the resulting handoff useful.

Built two conversation flows, tested three positioning variants, and generated structured sales-ready handoffs.

The shorter flow won, shifting the product toward a qualification layer before human sales rather than an AI replacement for sales.

A professional-services team needed repeatability, human review, and a safe integration path before production adoption.

Mapped the legal workflow, designed review gates, prototyped the AI-assisted flow, and defined the tool-aware integration architecture.

Created a measurable pilot path with accuracy, review time, adoption, and exception handling as explicit success criteria.

Fractional ownership had to be explained and sold inside a single ad, to people who assume property investing needs a large budget.

Ran Meta on UK audiences with static creative only, after it outperformed the alternative, and built the winning ad around the entry price itself: invest in UK property without a large budget, starting at £100 in fractional shares.

33 website leads at £1.93 on £63.71 of spend and a 4.69% all-click rate. One static visual out of the set carried 21 of those leads at £1.60 each, so the creative question was settled before scaling budget.

The offer was new enough that psychotherapists, psychologists, and clinics had to understand an unfamiliar treatment method before they would leave their contact details.

Ran twelve ad sets on Meta against US practitioners and clinics, splitting static feed creative against video that showed a session and positioned VR hypnotherapy as an addition to an existing practice rather than a replacement for it.

46 leads on $765.77 of spend — every one of them from a video ad set, the best at $10.12 per lead, while the static feed sets produced none. Blended cost per lead was $16.65 against a $150 monthly subscription.

Discovery, trust, delivery, checkout, and shop operations had to work as one local-commerce experience.

Designed both marketplace sides: occasion and location discovery, florist profiles, catalog, favorites, checkout, and order management.

Created a complete buyer-to-florist product flow ready for implementation and market testing.

Physical devices, presence and access control, and computer vision had to be translated into a coherent technical roadmap without coupling unrelated product functions.

Reviewed architecture and delivery for charging and access-control products, including neural-network number-plate recognition, and translated product constraints into technical and hiring priorities.

Established an implementation roadmap with clear boundaries between connected devices, access control, and computer-vision integration.

The pool needed a reliable internal redesign across Dash, BTC-GPU, Ethereum, Litecoin, and other currencies.

Reworked the NOMP-based platform and its operational architecture for multi-currency mining workloads.

Supported more than 5,000 connected mining devices across a multi-currency pool.

Monitoring, pool control, efficiency, and stock management were fragmented across a distributed physical operation.

Designed, developed, and implemented the software management system spanning facilities, devices, pools, efficiency, and inventory.

Created a unified operational view for site operators and owners of hosted mining equipment across facilities, devices, pools, efficiency, inventory, and customer dashboard functions.

After iOS 7 disrupted the early app for roughly one in three users, the product needed a full mobile rebuild, scalable backend and data processing, and a delivery team.

Designed and prototyped the architecture, implemented redundant high-load infrastructure, rebuilt iOS and Android apps, assembled the team, and supported business development.

Reached 200,000 users in the first three months; later public reporting recorded more than 500,000 downloads and about $450,000 in seed investment after a three-month IIDF accelerator.

The team needed evidence from both sides of a financing marketplace before building a broad creator platform.

Mapped segments, tested four offer framings, built an application-style MVP, and ran creator and capital-side conversations.

Narrowed the opportunity to creators with measurable recurring income and defined underwriting as the next proof gate.

General interest in automation obscured which workflow was urgent enough to adopt and pay for.

Mapped operational pain, selected a narrow workflow, built a human-reviewed functional prototype, and tested positioning by segment.

Proved workflow feasibility and clarified the segment, leaving a live-team pilot as the next evidence gate.

The program needed applications from founders it could actually accept, and the early creative rounds were buying leads at roughly twice the cost the economics allowed.

Ran Meta as the only channel across two audience sets — seed-accelerator intent and broader startup-company intent — and tested creatives, headlines, and offer angles until one held: address non-technical founders directly and promise they can stay on vision and fundraising while the program handles the rest.

74 website leads at $7.89 each on $583.97 of spend, with cost per lead halved over the test cycle. The seed-accelerator audience delivered leads at $6.48 against $10.09 for the broader startup audience, which decided where the budget belonged.

Demand had to be captured from clinics already looking for a telemedicine solution, which made search intent — not audience targeting — the thing to get right.

Built the keyword set from scratch, wrote the ads, and ran continuous keyword and campaign optimisation across two ad groups: one on telemedicine intent, one on broader digital-health wording.

20 conversions at €288.87 from €5,777 of spend at a 5.2% click-through rate. Telemedicine intent converted at €237.12 against €366.49 for digital health, and against a €15,000 average deal both stayed far inside what the business could pay.

Recipes from established authors and food blogs needed one simple mobile journey rather than a dense catalog.

Designed the product concept, UX architecture, discovery filters, favorites, shopping lists, and step-by-step cooking flow.

Produced an end-to-end mobile product experience spanning inspiration, planning, and kitchen use.

Photos, video, music, and 360° media needed creation mechanics built around group authorship rather than solo posting.

Defined the product strategy, creator UX, invitations, collaborative editing, media tools, and publishing mechanics.

Delivered a coherent social-product concept for creating and publishing stories together.

Turn an ambiguous human signal into a testable face-and-video analysis workflow without claiming a validated empathy model.

Designed a webcam-based capture and ML/CV analysis architecture for retail interactions and documented potential store, bank, and call-center scenarios.

Produced a presentation-ready architecture and proof concept; no trained or deployed production model was claimed.

Location-aware play, AR/VR, collective mechanics, and branching quest progression had to work as one coherent mobile system.

Designed location-bound quest flows with POIs and check-ins, resources, battles, sequential and branching steps, quest chains, and rewards.

Reached a working closed beta with more than 100 testers in December 2020; a later public release was planned but is not independently confirmed.