ENGAGEMENT MODELS
AI compresses portions of engineering work. Our commercial model reflects that: commit to a defined product milestone, fund the delivery foundation, then tie a meaningful balance to accepted engineering evidence.
SEVEN WAYS TO WORK
Commercial: fixed and prepaid.
Use when: product intent, feasibility, architecture or rescue scope is still ambiguous.
Leaves behind: baselined requirements, executable performance model, partition scorecard, risks, verification strategy and milestone SOW.
Commercial: fixed foundation fee plus a deferred tranche per evidence gate.
Use when: the outcome is important, bounded and objectively testable.
Leaves behind: accepted system capability, reproducible evidence and no single oversized project-end holdback.
Commercial: fixed program-phase fee covering a named backlog, delivery cadence and quarterly evidence gates.
Use when: architecture, RTL, firmware and lab priorities must converge across a sustained product phase.
Leaves behind: accepted increments and predictable access without converting engineering into an activity ledger.
Commercial: fixed fee per governed product phase with evidence-gate bonuses.
Use when: a team needs senior ownership of architecture, partition, reviews and verification sufficiency.
Leaves behind: governed decisions, risk disposition and a stronger internal team.
Commercial: setup, operating and acceptance-based transfer fees.
Use when: CI, regression, emulation, prototyping or agentic workflows must become durable internal capability.
Leaves behind: stable operations, documentation, training and transferred ownership.
Commercial: paid definition and a meaningful cash foundation, with a capped portion of earned fees deferred to an objective evidence gate or a defined financing event. Equity or options can be considered separately.
Use when: a selected funded startup must preserve runway while closing a consequential, objectively testable product gate.
Leaves behind: accepted engineering evidence and aligned economics without using equity as a substitute for delivery cost. Deferred terms require written legal and tax review.
Commercial: billed by the day or hour against an agreed rate card.
Senior FPGA and embedded engineering delivered against an agreed rate card, billed by the day or hour with transparent timesheets. No fixed scope or milestone commitment.
Best for: open-ended or exploratory work, staff augmentation, and support where the scope is expected to change.
This is the traditional consulting model.
A SPECIALIST EXTENSION OF YOUR TEAM
Your team retains the product context, technical authority and long-term roadmap. Magic Lens adds focused systems, FPGA, DSP/SDR, firmware, emulation and validation depth around a defined outcome and a shared acceptance process.
The model complements the internal team. It concentrates scarce cross-disciplinary depth on the immediate product constraint while keeping decisions, artifacts and operating knowledge in the client environment.
A named client owner remains inside architecture decisions, reviews and acceptance. Shared repositories, traceable decision records, reproducible builds and paired technical reviews make progress visible throughout the engagement.
Each gate leaves usable implementation, tests, evidence, runbooks and trained internal owners. Support can continue through a fixed runway or a measured build, operate and transfer plan.
Acceptance includes architecture decisions, interfaces, build/release manifests, tests, runbooks and transfer sessions.
Named client owners, review cadence, repository and CI integration, decision logs and explicit dependency dates.
Pre-existing methods, project IP, generated artifacts, third-party components and client data are separated in the SOW.
Two-to-six-week evidence gates expose progress early; dependencies and changes trigger documented rebaseline instead of silent drift.
Choose transfer, a fixed support runway or build, operate and transfer with measurable readiness for client ownership.
No. The engagement augments the existing team around a bounded delivery objective. Client leaders retain product authority, roadmap ownership and final acceptance.
THE EVIDENCE CONTRACT
Every risk-share begins with a one-page commercial evidence contract attached to the technical baseline.
Versioned input configuration, target metric, tolerance and test environment.
Named client owner, review window, test-backed rejection and cure path.
Client inputs, hardware, licenses and third-party services with owners and due dates.
New requirements or unavailable dependencies trigger extension or a fixed-price rebaseline.
Cap on deferred balance, milestone-level ledgers and a long-stop payment date.
Pre-existing and project IP, generated artifacts, models, data boundaries and third-party licenses remain explicit.
EXAMPLE ENGINEERING GATES
Measures are limited to what Magic Lens can materially influence. Revenue, fundraising, certification and supplier schedules are not engineering acceptance tests.
Reproducible build; lint and CDC/RDC disposition; regression and coverage rationale; timing, resources and target-board behavior.
Repeatable boot; API and driver conformance; bounded latency; error recovery; hardware-in-loop regression and release artifacts.
Reproducible compile/map; usable transactors or virtual peripherals; firmware boot; workload execution; regression throughput and debug observability.
Interface integrity; sustained-rate and fault testing; power or thermal measurements where in scope; open-defect disposition and knowledge transfer.
WHY THIS MODEL NOW
Current services research points toward domain-specific workflow ownership, hybrid outcome economics and measurable evidence. Hardware programs need the same shift, with tighter control of technical causality.