loop running · since 2019 at Carta

Mark Meleka

I write the loops that run the agents.

Software engineer, San Francisco. I run on a loop myself: get curious, go deep, ship, repeat — equity compensation, AI agents, healthcare, anti-spam, ad-tech, and counting.

Loop throughputTop of eng orgby merged PRs — agents do the work
Blast radius1000s of companiesmillions of equity records
Eval depth540-run RAG studyOxford MSc, 2025

Execution log — selected iterations

loop 2025–26
Put AI agents into Carta's products.
actBuilt Carta's first agent that works across its products — a hackathon MVP on internal Model Context Protocol (MCP) servers, pitched directly to the CEO.
eval✓CEO-backed, multi-team initiative
✓Production agent infrastructure — observability, CI/CD, secure agent-to-service auth
✓Created & lead org-wide “Working with Agents” trainings; AI on-sites with Carta's largest customers
loop 2023–25
Make LLM systems dependable, not just impressive.
actOxford MSc dissertation — built and evaluated a retrieval-augmented (RAG) assistant over health records; a 540-run factorial study.
eval✓Finding: reasoning, not retrieval, is the bottleneck on hard queries
✓Failure-mode taxonomy, now applied to production agents at Carta
✓Peer reviewer for digital-health journals (JMIR)
loop 2021–25
Ship across the platform, not one product.
actTotal Compensation, Cap Table, LLC, Stakeholder Management — plus the HRIS layer that feeds them all.
eval✓Two migrations off PII-heavy legacy data models — zero regressions
✓Org-wide API migration across six major versions — zero downtime
✓Equity-compensation domain depth — taxation, plan design, valuations — rare among engineers
loop 2019–21
Turn data into a product.
actBuilt the prototype that became Total Compensation; presented to executives weekly as the team formed around it.
eval✓Now used by thousands of companies across millions of employee records
loop 2014–19
Learn where it hurts.
actAnti-spam ML at Kik (tens of millions of users); ad-tech data at Sortable; WordPress sites end-to-end, solo.
eval✓Spam reports down ~70%, false positives held flat

Play

./originSmall-town Canada; development economics → software. Optimizing for how complex systems affect human outcomes.
./sidequests40+ countries · competitive debate
./researchAn illustrated, cited long-read — how China grew, who gained, and what comes next.
./lightsA classic pattern game, rebuilt from scratch. Play it, then read the linear-algebra solution Grigory Solomatov and I wrote.
loop next
Trade notes with people building hard things.
observeYou, reading this.
planA conversation.
act[email protected]
evalPending▌· type to reply