Master orchestrator fires every 30 min from 8am–7pm IST and routes to eight sub-workflows via webhooks. Each runs trigger → config → GPT-4o agent (temp 0.3) → output formatter. Outreach lands in Gmail drafts, research in Notion, spend lives in a USDC wallet.
The actual node graph from my n8n — rendered live. Retell posts the transcript to a webhook, we respond 200 immediately, then a LangChain agent extracts name, firm, disbursement, disposition. Two IF gates route interested, callback and not-interested to the right tab in the leads sheet. Zero human in the loop until follow-up.
Wheel bot. Wakes every 15 minutes during US market hours, queries Alpaca options data, picks a 14–28 DTE NVDA contract near 0.25 delta, filters on IV rank ≥ 30, and writes the trade. A 2x premium max-loss cut and 0.70 delta roll keep the tail behaved. Last closed cycle: sold NVDA260424P00180000 for +$206.92 net (after OCC + ORF + TAF fees), held 14 days to expiry, on $18K of collateral. Compounded across the 14-day cadence, that pace projects to a ~18.8% 7-month rolling ROI if the cycle repeats cleanly.
Congress mirror. Pulls Periodic Transaction Reports (the STOCK Act filings) and mirrors meaningful trades into an Alpaca portfolio, using capital.com market data for execution prices. Filings publish with a 30–45 day lag by law — so the strategy is a slower, signal-driven tilt, not a reaction trade. Currently paper-only with a 7-day cooldown post-filing.
Caveats: paper-trading until Sharpe + max-drawdown clear my thresholds. The 18.8% figure is a cycle-repeat extrapolation, not a track record — IV regime, assignments, and roll losses will compress it. Capital.com data is delayed for non-pro tiers; size positions accordingly. Nothing here is investment advice.
Daily SLA monitor scans every active outreach card at 9am and flips SLA-breach checkboxes. Stage-movement router fires on any kanban card move — stamps last-activity, resets SLA, sets the next action. Weekly report generator calls Claude Haiku to produce three audience-specific reports — founders, VP, manager — and writes them into Notion on Monday morning. Whole stack runs on roughly $25/mo of AI tokens.
Fixed viewport on purpose — terminals aren't responsive. Four quadrants: markets and FX, OSINT news and events, the DeckGL globe with stacked layers (exchanges, conflict zones, sanctions), and an analytics pane. A scrolling ticker bar across the top because of course it does.
A master orchestrator skill routes to fourteen sub-skills that each grade one dimension of a site. Output is an interactive HTML dashboard with a radar chart, per-dimension scores, and prioritized fixes. Pairs with a frontend-design skill for UI and accessibility read-outs.
End-to-end autonomous systems. n8n hub-and-spoke orchestration, MCP protocol for tool use, Retell AI for voice, Ollama on-box when latency or privacy matters. I think in workflows, not prompts.
Python bots on Alpaca. Options math, risk sizing, idempotent schedulers. Paper-first, Sharpe-gated.
DeckGL terminals, Power BI for the enterprise side. Bloomberg aesthetic, event detection.
Audits that actually score things. SEO, CRO, positioning — interactive dashboards instead of 60-page PDFs.
Obsidian vault that self-improves with Claude. Karpathy-style hub-and-archive, brain score tracked.
Cross-domain context — trading, voice AI, OSINT, marketing. I ship the thing then write the doc.
I'm Shayanto, a builder in Bangalore who treats the product/engineering/growth boundary as a polite suggestion.
I ship AI agents, voice systems, trading bots and growth audits — often the same week — by using AI the way engineers use libraries: as leverage.
Looking for a seat where generalism is a feature, not a bug. Product, growth, founder's office, applied AI, solutions.
A local MCP server + rules.json that lets Claude read chart state and resolve setups via a context router.
Pine Script v5 indicator for a 5-min opening range + 1-min breakout with a 1:2 R:R frame.
A self-improving vault with Karpathy-style consolidation loops and a brain-health metric.
A dozen custom skills for audit, trading, process automation, and knowledge-ops rituals.
Private marketplace of Claude plugins scoped to product, growth, marketing workflows.
Working voice agent demo built in a weekend to answer an application with a thing, not a deck.
Outbound concierge agent for Divyasree's Private Valley villa plots near Nandi Hills. 4 qualification checkpoints, EN+HI, sub-3-minute calls. Built on Retell AI.
Product-design assignment: a fintech onboarding dashboard that makes the path to go-live fully transparent. Shipped prototype in Satoshi + JetBrains Mono.
71/100 full-suite audit — CRO and schema gaps flagged, content depth prioritized. Portfolio case.
Hiring for product, growth, founder's office, or applied AI? Pick a channel — I'll bring 20 minutes and a sharp opinion on whatever you're stuck on.