Joseph C.
Forward Deployed Engineer · GTM Engineer · Sales Engineer
Remote · I put AI systems into real customer workflows and stay until people use them
Play the video · from customers to code: the path, BDRclaw, MyGrokFlow, TickerFront, PumpWire · no sound
I have sat on both sides of the deal: carrying a quota-adjacent seat (sales engineering, account management) and shipping the code that the customer actually runs. That is the job description of a Forward Deployed Engineer.
| If you need a... | What I bring | Proof below |
|---|---|---|
| Forward Deployed Engineer | Scope with the customer, build inside their stack, hand over something their team runs without me | PumpWire, MyGrokFlow, TickerFront |
| GTM Engineer | AI outbound and lead systems: research, enrichment, gated drafts, reply triage, CRM as the source of truth | BDRclaw, MyGrokFlow |
| Sales Engineer | Demos and POCs that survive technical buyers; I translate product to buyer and back | SE + account management background, AgentToll and PumpWire x402 demos |
Full resume on request through josephdev.online/resume.
I build under deadline and ship the proof, not the slides: agent payments, agentic finance and multi-agent pipelines.
|
COLOSSEUM · CRYPTO WORLD'S FAIR 2026
Charge AI agents per request. Keep humans free.
|
PAYPAL AI HACKATHON · DEVPOST 2026
An AI back office for businesses that run on PayPal.
|
|
ANSEMHACK CLAWRENA · CLAWPUMP × PUMP.FUN
pump.fun rug-risk scores, sold to other agents per call.
|
LABLAB.AI · AI AGENTS HACKATHON
Tweets that write and post themselves.
|
-000000?style=flat-square&logo=x&logoColor=white){kind=icon}
PumpWire, explained simply (click to expand)
- The problem. pump.fun lets anyone launch a Solana token in seconds on a bonding curve (the price rises automatically as people buy). That speed also makes rug pulls easy: a developer launches, friendly wallets buy first, then everyone dumps on later buyers. A trading bot looking at a brand-new token can't easily see who deployed it, who bought first, or whether those early buyers are secretly the same person.
- What PumpWire does. It watches every pump.fun launch onchain, stores launches, trades and wallet funding links, and turns them into a 0 to 100 rug-risk score with plain-English reasons (e.g. "3 buyers in the creation slot share one funding wallet"). The score is a deterministic, versioned, unit-tested function, so every number traces back to its evidence.
- How agents pay for it. No API keys or subscriptions. It uses x402, a revival of the web's
HTTP 402 Payment Requiredstatus code: an agent callsGET /v1/risk/:mint, gets a402with the price ($0.01 USDC), pays on Solana, retries, and gets the answer. - How agents plug in. It ships as an MCP server, so Claude, Cursor or a claw-agent can simply ask "rug check
<mint>" and pay from their own wallet, with spending caps enforced before anything is signed. - Proof, not slides. Our own buyer agent (PWIRE Scout) made 250 paid mainnet calls on launch day. A public
/livepage and/v1/statskeep first-party calls separate from third-party ones, so the usage numbers stay honest. - Stack. TypeScript/Node workers → SQLite (WAL) → Express + x402 API → MCP client, run on my own VPS.
$PWIRE mint: 2b2Tv315U1FUtYF9Y1H2b2qrCnL3tN5QPabScFPCw8vw · #AnsemHack · Risk information from public onchain data only, not investment advice. $PWIRE is a utility token for the hackathon.
|
Pair arbitrage on Polymarket's 5-minute crypto markets. Rust on the hot path, Python for research, an explicit order state machine. Cut identify-to-send latency from 500 ms to 130 ms by measuring first and moving regions. |
High-quality investor-relations websites for small publicly traded companies, plus a scraper that finds the ones that need it. The site an investor lands on before reading the filing. |
|
GTM IAAS for a client that I built. Helps GTM Agencies and the GTM process become optimized from a cost reduction standpoint and increase scalability. |
An AI BDR: prospect research, outreach drafts behind a human approval gate, reply classification, meeting booking and CRM sync across email, SMS and LinkedIn. Built to show safe agent design where AI touches customers. Walkthrough on request. |
|
My site in four languages (EN / ES / PT / IT): an About page with the 90-second intro video, and a resume you request through a form that routes straight to my inbox. |
|
- Scope and price in writing before a line of code. The customer approves the brief; I send progress reports while I build.
- Keep the model out of the hot path. Deterministic code makes the decisions that move money or touch customers; the LLM researches, drafts and reviews.
- Harness before horsepower. Evals, quality gates, spend caps, kill switches and a human approval step before any risky action.
- Hand over, don't hold hostage. Docs, runbooks and a team that can run it without me. Ship something an account team can defend in a QBR, not a demo that dies after the call.
Global consulting firm (client operations) → Sales Engineer, B2B SaaS → Account Manager, B2B SaaS → independent Forward Deployed Engineer shipping AI systems for operators and my own companies. About six years in total, all remote.
That path means I can own account health, renewals and escalations, run discovery and technical demos, and then build the thing myself.
Education: M.S. IT Management (WGU) · B.S. Finance (UB)
Customer & GTM · discovery · demos/POCs · solution design · account management · renewals and expansion · CRM discipline · stakeholder communication
AI systems · agent workflows · MCP servers · RAG · evals and harnesses · human-in-the-loop gates · x402 agent payments
Build · TypeScript · Python · Rust · Node · Next.js/React · SQL/SQLite · APIs · Cloudflare Workers · Linux VPS ops · Git
Languages · English (native) · Spanish (B2, professional) · Italian (B1) · Portuguese (B1)
Open to remote Forward Deployed Engineer, GTM Engineer and Sales Engineer roles · request my resume