Prince Jain

I build and operate production systems at Adelev8 (growth marketing and venue tech): product, code, infrastructure, payments, and security across a five-node Linux fleet.

venue os - the thing i run

Adelev8's venue-operations platform runs a 30,000 sq ft entertainment park in Bengaluru: direct-ticket site, self-service kiosks, cashier POS, ESC/POS printing, QR turnstiles with anti-passback, Razorpay UPI reconciliation, vendor split payouts, rain-day ticket transfers, server-side Meta CAPI.

• 10,238 recorded transactions across 3,526 paid bookings and 7,116 POS orders
• 170K+ tracked visit events, 46 applied Postgres migrations
• peak day at 994 tickets and 937 gate check-ins

the fleet i operate

Five self-operated nodes: 26 CPUs, ~105.7 GiB RAM, 51 docker containers, 49 PM2 services, nginx across ~155 hostnames. Node roles: the Adelev8 box (venue, Saroj Jain, Quantiti), the pharma client box (SPL Life Sciences, Saroj Jain CRM, Plane, OpenSEO), the Hostinger box (MyToyTree's 8-container ERP, Livetree, self-hosted mail), an Oracle Ampere agent node, AWS EC2. Plus a self-built health dashboard over all of it.

client systems i host and built

• MyToyTree - toy retail: full ERP stack, POS, invoicing, self-hosted mail (Stalwart + Roundcube)
• Livetree - ops platform with an HDFC-bank integration service
• SPL Life Sciences - pharma systems on a dedicated node
• Saroj Jain - D2C brand: commerce stack, iOS app, 8-service self-hosted CRM (5x first-month growth, 2.71 ROAS on Myntra)
• Meghdoot Textiles - custom CRM, ERP, HRMS + WhatsApp retailer automation
• Payments clients - Suparna Ghosal (dedicated Razorpay service), Indian Souvenir Store, Garlic Brothers Exports, Credence Wire, Trusser
• Brands - BMW KUN Exclusive, Mulveer Jewellers, Ananta Karunai, Stones Gallery, 2thewall

other work

• Quantiti - Quantitative Software Engineer. Python quant platform routing real orders to Zerodha: ~Rs 1.5 Cr notional in 4 weeks across MCX, equity and F&O, kill-switches and per-fill slippage tracking. Research on self-orchestrated GPU (Kaggle T4s, ephemeral Vast.ai RTX 3090s that self-destroy).
• Maninfini Automation - Software Engineer & Platform Consultant (Partner). B2B SaaS and automation for pharma and jewellery houses.
• Independent - seven live KDP titles, seven App Store Connect records (four public), licensed ElevenLabs voices, contractor at Mercor.

open source

npm CLIs & tools: agent-trials (run claude/codex/gemini through one task suite, scored by their own verify scripts), ctxvitals (token cost of MCP servers, skills, rules), compactsim (counterfactual autoCompactWindow replay & Context Rent profiler for Claude Code and Codex logs), mcp-bouncer (stdio MCP policy proxy), canimcp (MCP client quirks preflight), slopsieve (AI-slop diff reviewer), uislop (frontend vibecode linter), agentblast (agent blast-radius & SSH auditor), egressleak (outbound HTTP egress & User-Agent PII leak auditor), repotrap (prompt-injection, invisible Unicode & repo-open trap scanner), loopbrake (agent edit-thrash, 39.2% shell-write bypass & doom-loop breaker), clonesieve (structural function clone, duplicate helper & dead-export auditor + 39.0% blind-write transcript scanner), ruleherd, mcp-tax, dep-census, readmeroast, and MaintainerOS.

Upstream (38 merged PRs): merged fixes in Arize-ai/phoenix (#17816, #17818), tokio-rs/bytes (#862, #863), seanmonstar/reqwest (#3118), agno-agi/agno (#11526), apache/datafusion (#26551), apache/arrow-rs (#11845), lancedb/lancedb (#4584, #4586), oxc-project/oxc (#28008), sqlalchemy/sqlalchemy (#13710, merged via Gerrit), pmd/pmd (#7225), connectrpc/connect-es (#1849), JSQLParser/JSqlParser (#2838 through #2843), redis/lettuce (#3966), Pebble (#842, #843, #845, #846, #847 - #843 fixed a comparison bug open since 2021), Handlebars.java (#1200, #1201), elie222/rakazo (#2125 through #2138, 8 merged PRs), and utk09-NCL/skardela (#12, #13). Fixes and regression suites I authored in vercel/ai (#23935) and NousResearch/hermes-agent (#150992) shipped with Co-authored-by credit. In review: browserbase/stagehand (#3441), vercel/ai (#24764), inngest/inngest-js (#1880, #1882), BerriAI/litellm (#50674, #50662, #50657, #50655, #50645), agno-agi/agno (#11549), Comfy-Org/ComfyUI (#17617), Arize-ai/phoenix (#17746), pydantic/pydantic-ai (#12864), triggerdotdev/trigger.dev (#5105, #5052), vercel-labs/agent-desktop (#624, #610, #635), petgraph/petgraph (#1066), axios/axios (#11362, #11363, #11367), oxc-project/oxc (#28109, #28110), ray-project/ray (#67229), msgspec (#1302), TanStack/table (#6639), and NousResearch/hermes-agent (#146938, #146940, #151383). 1,250+ contributions in the last year.

git + agents

Parallel agent runs on real worktrees, branches and rebases are my daily shape of work. I benchmark the agents (agent-trials), audit their config files (ruleherd), meter their context cost (ctxvitals), gate their MCP tools (mcp-bouncer, canimcp) and lint their output (slopsieve). Making a codebase agent-ready is most of what I already do.

reach me

github.com/princejain756
w.prince.sh/website
maintaineros.prince.sh
linkedin · x · [email protected] · +91 8005634678