Makhanikhi - APP

Makhanikhi is an Alexa+ MCP voice copilot empowering informal mechanics with hands-free AI diagnostics, visual catalogue cross-referencing, and instant local spares ordering under the chassis.

Comment

Makhanikhi: Voice-Assisted On-Chain Escrow & Logistics for Mobile Mechanics

What Inspired Me

I experienced the problem firsthand. First as an out-of-warranty bakkie (pick-up) owner unable to afford main-dealer service rates, and later by actively shadowing and apprenticing alongside a local mechanic in Polokwane townships. While working under the vehicle chassis in grease-covered coveralls, I saw mechanics struggle with a fundamental physical barrier: you cannot safely type, look at a smartphone screen, or search for parts when your hands are covered in oil. In high-density South African townships, millions of out-of-warranty vehicles depend on informal mobile mechanics. However, this informal economy operates under severe structural friction:

  1. Client Vulnerability: Vehicle owners face zero repair transparency, deposit fraud risk, and untracked second-hand parts.
  2. Mechanic Risk: Skilled mechanics do heavy physical labour without guaranteed payment, facing client default or delayed call-out fees.
  3. Operational Hazard: Mechanics lose billable time wiping hands to order parts, log repair steps, or calculate quotes on mobile screens. I built Makhanikhi to replace this high-friction "gig trap" with a hands-free, voice-activated, escrow-backed protocol that protects payments, verifies genuine parts, and builds an immutable digital resume for mechanics.

How I Built It

Makhanikhi combines Amazon Developer technologies, decentralised finance infrastructure, modern Web2 frontends, and AI workflows:

  1. Hands-Free Chassis Voice Agent (Alexa+ Track): Built a custom Alexa+ Agent Skill that enables mechanics to query parts compatibility, log telemetry steps, and trigger escrow releases entirely through voice prompts while working under the car ("Alexa+, ask Makhanikhi to record oil filter replacement and check ZARU escrow balance").
  2. Cloud Infrastructure (AWS Builder Challenge): Deployed backend microservices on AWS using AWS Lambda serverless functions and Amazon DynamoDB for real-time state synchronisation between Alexa+ voice intents, web clients, and on-chain program logs.
  3. Smart Contracts & Escrow Vaults: Built using Rust on the Solana blockchain (Anchor framework), leveraging Program Derived Address (PDA) vaults to lock Rand-backed SPL stablecoins (ZARU) for sub-cent micro-settlements.
  4. Frontend & Mobile Access: Developed as a responsive web application hosted on Vercel, integrating @solana/wallet-adapter for client and mechanic wallet authentication.
  5. Immutable Logs: Vehicle maintenance records and SAPS-compliant parts serial numbers are pinned off-chain via IPFS to create an immutable digital logbook for each vehicle.
  6. Generative AI Copilot: Accelerated development using Google AI Studio (Gemini) to prototype state-machine logic, map Alexa+ agent slot intents, and generate robust test suites.

The Economics & Math Behind Makhanikhi

To ensure trust without introducing predatory platform fees, Makhanikhi uses a dual-sided micro-fee settlement model for every completed job $J$:

Let $V_J$ be the agreed total job value locked into the PDA escrow vault. The net mechanic payout $P_m$ and client cost $C_c$ are defined as:

$$P_m = V_J \cdot (1 - \phi_m)$$$$C_c = V_J \cdot (1 + \phi_c) - R_{co}$$

Where:

  • $\phi_m = 0.025$ ($2.5\%$ mechanic service fee deducted upon instant settlement).
  • $\phi_c = 0.025$ ($2.5\%$ client platform facilitation and dispute protection fee).
  • $R_{co}$ is the Co-Pilot Inspection Rebate ($10\%$ rebate unlocked when the client inspects work and signs off on step-by-step repair telemetry):

$$R_{co} = 0.10 \cdot V_J$$

This formula mathematically aligns incentives: mechanics are guaranteed immediate liquidity upon job sign-off, while clients are rewarded for verifying repair progress.

Challenges I Faced

  1. Hands-Free Workflow UX in Noisy Environments: Developing the Alexa+ agent voice interaction required designing fault-tolerant voice intent slots that could parse the mechanic's commands despite ambient driveway background noise (power tools, roadside traffic).
  2. Real-World Operational Disconnects: Field testing with local mechanics revealed deep-seated friction around call-out fees, fuel costs, and informal financial management. These dynamics required pivoting protocol rules away from loose verbal agreements toward strict automated escrow triggers, Alexa+ voice confirmations, and transparent cost-calculator rules.
  3. Web3 Abstraction: Abstracting Web3 wallet complexity for township drivers and mechanics so the underlying PDA vaults operate silently behind standard ZAR stablecoin balances.

What I Learned

Fieldwork taught me that technology alone cannot bridge a trust deficit, because it must fit the physical realities of human labour. By integrating Alexa+ voice capabilities, I learned that true accessibility for field technicians means removing screen dependencies altogether while keeping payments secure and transparent.

Built With

Share this project:

Updates

Submission history