Systems we've shipped — and what they automate.

Manual B2B outreach doesn't scale for a solo founder. Four coordinated n8n workflows replace it.

• Lead Discovery — Apollo.io filters by ICP, dedupes against Supabase
• Invite Sender — Unipile API sends 20+ personalized connection requests daily
• Follow-up Sequencer — multi-touch cadence after acceptance
• Reply Monitor — Gemini classifies replies, routes hot leads to Telegram

Print-only tenders, property transactions, and business signals in Gujarati newspapers are invisible to digital tools. BidAssist and TenderTiger only cover digital eProcurement — they miss print entirely.

• WF1 — downloads Sandesh ePaper, 39 editions × 16 pages daily (624 pages)
• WF2 — Gemini Vision OCR extracts 5 signal types per page
• WF3 — serves structured alerts: tenders, property, business openings, court records, political appointments

3,500+ video corpus from a single creator with no way to query it conversationally. Built a persona-tuned hybrid retrieval layer.

• Transcript scraper → 1,224 videos ingested
• Chunking — 400 words, 50-word overlap → 3,500+ chunks
• Embedding — gemini-embedding-001, 3072 dimensions
• Retrieval — hybrid BM25 + semantic on pgvector
• Inference — persona-tuned LLM interface

B2B cold email at scale needs coordinated DNS, warmup, lead sourcing, and AI copywriting. Most founders cobble it together wrong.

• SPF / DKIM / DMARC setup and Mailreach warmup
• Apollo lead pipeline → NeverBounce validation → Supabase
• n8n orchestrates AI copy generation (Gemini) per persona
• Instantly / Smartlead delivery with reply detection

Lawyers and businesses miss hearing dates from eCourts. No proactive alerting exists — this fills it.

• Case status polling across 600+ courts via API
• Supabase state tracking for diffing
• WhatsApp / Telegram alerts on hearing date or status change

Async B2B automation engagements for SaaS founders and SME ops leads with repetitive manual processes.

• Intake form → diagnosis call → workflow delivery → documentation
• $299–$999 per workflow scope
• Async-first: no standing meetings, fixed scope per engagement

Architecture and build of multi-agent systems for teams that need agentic infrastructure without a full ML team.

• Orchestrator + subagent topologies
• Persistent memory layers (pgvector + structured stores)
• Tool-use agents with structured output pipelines