PROJECTS

Project showcase

A selection of representative engagements and projects from past work. Company names, project names, and identifying details are omitted for confidentiality.

FINTECH · AI PRODUCT & GROWTH

AI-enabled FinTech platform, 0→1→scale

Situation.: A multi-product FinTech portfolio with weak engagement, unclear positioning, and spend that wasn't converting — several products competing for attention without a clear core user.
Intervention.: Defined four core personas and refocused the roadmap and messaging around the highest-value ones. Rebuilt acquisition and activation around those personas, tightened onboarding, and ran a weekly experiment cadence with engineering, data, and design. Owned the product portfolio end-to-end — PRDs, roadmap, and growth — and built and managed the team delivering it. Across the portfolio this included a perpetuals investment platform and an AI-assisted investing experience that profiles a user's risk posture to propose suitable assets — designed to remove complexity for investors in their late 30s–50s who are interested in non-traditional securities but are not crypto-native.
Evidence.: ~2.5× user growth and doubled revenue; +55% engagement and 40% shorter time-to-market.
Capacity: Embedded product & growth lead through an SDTR client engagement.
Team: cross-functional team of 8.
Responsibilities.: Owned product and growth for the portfolio: personas, roadmap, PRDs, experiment cadence, onboarding and activation, KPI design, and stakeholder alignment — including building, hiring, and managing the cross-functional team delivering it.

APPLIED AI · PRODUCT STRATEGY

Reframing a stalled AI product, and opening an enterprise path

Situation.: A consumer conversational-AI product built on a strong proprietary language model, but with stalled engagement and weak monetization — strong technology, unclear product-market fit.
Intervention.: Ran funnel analysis across onboarding, repeat use, and channel feedback to find where the product actually resonated. Reframed the consumer experience around a higher-retention use case, then originated a new B2B business line on the same engine — a knowledge/onboarding product using private retrieval — and designed its strategy and foundation, including the RAG approach and an evaluation rubric for answer quality, latency, and cost.
Evidence.: On the consumer side, ~3× early-week retention and ~50% longer sessions in a matched cohort, within weeks. The B2B line went from proposal to PoC and a limited pilot, reducing mentor queries ~20% and time-to-answer ~65% while maintaining answer accuracy.
Capacity: SDTR consulting engagement.
Team: worked with the client's product and engineering teams.
Responsibilities.: Led discovery and funnel analysis; defined the retrieval and evaluation approach; wrote the product requirements; set the B2C→B2B direction and the metrics to steer it.

APPLIED AI · RAPID PROOF-OF-CONCEPT

Rapid proofs-of-concept for early validation

Situation.: Teams needing evidence before committing serious build — an AI opportunity on the table, stakeholders unaligned, and no working proof to decide with.
Intervention.: A repeated pattern across engagements: align with stakeholders and management on the question that matters, analyze the competitive landscape, then scope and build a working proof against it — conversational agents, retrieval-grounded knowledge assistants, staged model pipelines, AI forecasting. The same discipline is applied outside client work: SDTR regularly builds under competition conditions to pressure-test new frameworks, technologies, and product concepts with external validation — staying at the edge instead of reading about it.
Evidence.: Client PoCs that advanced to pilots and product decisions. A separate engagement paired an end-to-end product strategy — diagnosis, sequencing, monetization logic, and metrics — with a deployed working prototype, documenting implemented and simulated components separately. smartyPants, a real-time AI conversation copilot, began as the winning build at the Google Japan / DeepLearning.AI AI Voice Agents Hackathon and was subsequently launched as a product. Other competition builds also won four awards at ETHGlobal Tokyo.
Capacity: SDTR engagements and independent competition builds.
Team: solo or small cross-functional team.
Responsibilities.: Stakeholder alignment, competitive analysis, PoC scoping and build — product definition through working demo.

APPLIED AI · EVALUATION & RELIABILITY

A real-time conversation assistant with evaluation built in

Situation.: Useful live conversation assistance depends on timely, grounded outputs and controlled behavior when models or providers fail. A convincing demo alone cannot show whether a change improves the experience.
Intervention.: Built a private bilingual assistant combining transcription, translation, and contextual suggestions. Recorded-conversation replay exercises the live processing pipeline; comparative model evaluation and traceable records support model and prompt decisions. Output validation, regression checks, execution limits, and bounded provider-failure handling are part of the implementation.
Evidence.: A working private assistant with an implemented evaluation workflow and runtime controls, including checks for unsupported outputs and provider failures. Model and prompt changes are compared against the current version before adoption.
Capacity: In-house product development (private).
Team: Founder-led.
Responsibilities.: Product definition, system design, implementation, evaluation, runtime safeguards, and cost and latency controls.

CONSUMER / DTC · PRODUCT

Product invention for an underserved segment

Situation.: An established but conservative premium-goods category, dominated by heritage players, sold almost entirely offline through intermediaries. Little product innovation at accessible price points, and no direct relationship with a younger, digitally-native buyer.
Intervention.: Identified an underserved segment — high-earning, time-poor, design-driven buyers priced out of true personalization — and built a product line specifically for them: premium, customizable, sold directly online before DTC was a common approach. New products were created analytics-first: ecommerce, website, and advertising data fed directly into what got designed and made. The online channel itself was run as a product — multi-country sales, multi-currency payments, BNPL and crypto payment options — alongside a rapid manufacturing strategy (CAD/CAM, CNC, 3D-printed prototyping) to validate before committing to production.
Evidence.: Profitable in year one; 250% ROI by year two. A signature product sold out at launch and drove unpaid coverage in major international media; +50% awareness YoY, +25% sales, and development cycles cut from 18 to ~4.5 months.
Capacity: Co-founder and former CPO; led product and ran the business end to end (operating role).
Team: with a co-founder on manufacturing/craft.
Responsibilities.: Led product strategy and full go-to-market — segment definition, product concept, the eCommerce site and online customization tools, pricing, the DTC model, payment methods and sales channels, and data-to-product loops — and ran the business: P&L and finance, hiring and team management, marketing and sales.

ENGINEERING · DELIVERY LEADERSHIP

Internal tooling that scaled across a large organization

Situation.: A services organization running 40+ contact centers across Europe, each with its own tools and processes — onboarding a new client or center took two weeks, and inconsistency slowed delivery.
Intervention.: Led a small engineering team to build internal tooling that standardized the work: a live analytics dashboard fed by database replication, a drag-and-drop system for spinning up new phone-handling flows, and a campaign-configuration tool to cut rollout friction. Gathered best practices across centers and embedded them into the tools.
Evidence.: Cut new-client onboarding from ~14 days to ~2 and grew the CRM portfolio 30%; internal quality scores for new integrations rose sharply. The tooling was adopted into regional operating playbooks.
Capacity: Team lead (operating role).
Team: 7 engineers.
Responsibilities.: Led the 7-engineer team; designed the systems; owned delivery, rollout, and training across the centers.