Builder and operator · Building Luminik
12+ years shipping software · Oslo · B2B products · engineering leadership · founder-led sales
prasad.tech · LinkedIn · Email
Luminik is an event pipeline platform for B2B teams. It helps sales and marketing teams connect attendee sourcing, enrichment, sequencing, capture, and attribution in the systems they already use.
I am building the company hands-on: product, engineering, early sales, customer work, and the operating cadence around it. My bias is to sell before building, work alongside customers, and build repeated workflows into product.
The public engineering side of that work is Alfred, an open-source framework that runs fleets of coding agents on Claude Code and Codex. It runs Luminik's engineering every day; in a 30-day window (Jun to Jul 2026) the fleet merged 265 agent-authored PRs across 10 repos, with human review on every merge. Docs at alfred.luminik.io.
- Aura at Bain Founder's Studio: PE due-diligence software. Took it from 0 to
$3.6M ARRin15 monthswhile the engineering team grew from five to thirty-three. - Mainteny: co-founded the European field-service SaaS company. Built the initial product, worked directly with early customers, helped raise a
$2.7M seed, and helped build a high-performing product and engineering team. - SnowOptix: Snowflake cost optimization built on
ACCOUNT_USAGEsystem tables, using Snowflake Cortex and Snowpark Container Services, distributed as a Snowflake Native App. Entered Snowflake's Startup Accelerator, and the customer discovery behind it became Luminik. - BlueJeans: backend engineer on video conferencing, starting in 2014. My first deep exposure to a scaled engineering organization.
I like work where the distance between customer conversation, product decision, and shipped code is short. I have led teams as they grew from founding groups into larger engineering orgs. The parts I care about are clear ownership, direct customer context, strong engineering standards, and a steady product rhythm.
At Luminik, I stay close to the product and the sales work while an agent fleet on Alfred carries the day-to-day engineering: specs-driven planning, isolated git worktrees, automated PR review, and approval gates. I decide product direction, review every merge, and handle customer commitments myself.
| Area | What I stay close to |
|---|---|
| Product | Customer workflows, product shape, architecture, and release quality. |
| Sales | Founder-led discovery, objections, onboarding, and buyer follow-up. |
| Team building | Hiring, standards, ownership, operating cadence, and technical direction. |
| Agents | Specs, worktree isolation, PR review, evals, memory, and spend caps. |
| Writing | Public notes on AI enablement, agents in production, and zero-to-one work. |
- Alfred: open-source framework for running fleets of coding agents on Claude Code and Codex (alfred.luminik.io).
- prasad.tech: writing on AI enablement, agents in production, founder-led sales, and zero-to-one product work.
- Luminik site: public company site for the event pipeline platform.
While setting up the agent stack, I contributed fixes and repros to tools I depend on.
- Hermes Agent #12251: contributed to the Gemini auth regression fix.
- Hermes Agent #12224: contributed the WhatsApp formatting issue and expected behavior.
- gbrain #89: contributed testing notes on Gemini embedding support.
Long-form notes on AI enablement, agents in production, founder-led sales, and zero-to-one product work.
- A field guide to adopting AI in engineering
- Deploying a customer-support AI agent to production: testing it, then trusting the numbers
- Run AI enablement as a P&L
- Loops, harnesses, and memory: what it takes to run an autonomous coding fleet
- How to build an LLM judge you can trust
- Durable by design: two ways to survive failure in AI-native workflows
- Three levers for cutting your coding agent's token bill, and how to tell which one works
- Writing the verifier is the real work in agent RL
- Technical founders can sell: here's how
- Building a $3.6M ARR product inside a consulting firm
- Attributing Snowflake cost to the query that caused it
| Layer | Tools |
|---|---|
| Backend | Kotlin, Quarkus, Python, FastAPI, Postgres, AWS ECS |
| Frontend | React, TypeScript, Vite, Astro, Tailwind, shadcn/ui |
| Mobile | Expo, React Native, on-device ML (ML Kit, Apple Vision) |
| Data and integrations | Kafka, Snowflake (Native Apps, Cortex, Snowpark Container Services), Nango, CRM and sequencing integrations |
| AI and agents | Alfred, Claude Code, Codex, MCP, LangChain, LangGraph, Langfuse, DeepEval, RAG (pgvector), evals and LLM-as-judge, agent memory |
| Workflow orchestration | Hatchet, Temporal, DBOS, GitHub Actions, specs, branch isolation, release checks |
| Leadership and operating | Hiring, product planning, customer discovery, team cadence, technical direction |
| Writing | Technical essays, product notes, operator docs |
LinkedIn · prasad.tech · prasadus92@gmail.com
For Luminik: luminik.io





