Flight log · New York City

LUV JAIN

CALLSIGN: CURIOSITY // TECHNICAL PRODUCT MANAGER · FOUNDER OF CODEPOET

Software engineer with the soul of a mad scientist. Nine years flying through developer platforms, fintech, and one AI startup. This is the mission log: five planets, one chess puzzle, a rover with a message, and a shelf of stories. Poke everything. That's the point.

OPEN TO applied-AI roles · eng manager or technical PM · fintech · NYC or remote

9 yrsin flight
5missions
$300M+assets tracked
1startup live

short on time? grab the résumé · skip to contact

↑ a small solar system of my career. click or tap a planet to fly there.

scroll to launch ▾

Mission 01 · 2015–2016 · Seattle

Planet AmazonSoftware Development Intern · Alexa Machine Learning Platform

First landing. A jungle planet of graph databases and voice assistants, where I learned that shipping beats theorizing.

  • Built an event-based microservice ingesting prototype data into Blazegraph (the precursor to AWS Neptune), with a ~100ms query API over the Alexa ontology using RDF triplets.
  • Built an ML inference store for Alexa's knowledge database that later shipped in production Alexa Brain.
Graph DBsRDFML pipelinesAlexa
TRAJECTORY · INTERN ENGINEERlearned: real systems have real users. ship accordingly.

Mission 02 · 2017–2019 · McLean, VA

Planet Capital OneSoftware Engineer · Card & Small Business Technology

A ringed world of fintech at scale. Fraud platforms, event-driven everything, and my first taste of infrastructure as code.

  • Automated a PySpark pipeline processing 50,000 customer debt records daily, asynchronously, on cloud architecture.
  • Designed Terraform event-driven microservices to onboard Walmart customers onto a fraud-detection platform.
  • Shipped an API + Spark workflow enrolling 20,000 new accounts on AWS Lambda and S3. (Circle of Excellence Award.)
TerraformAWSPySparkMicroservices
TRAJECTORY · ENGINEER SENIOR ENGINEERlearned: when downtime isn't an option, boring is a feature. own the whole system, not just your commits.

Mission 03 · 2019–2024 · Seattle

Planet OracleSenior Software Engineer · Access Control Ecosystem

The red planet. Four years of government-grade identity infrastructure, where 99.99% uptime is the job description and downtime is not an option.

  • Architected OCI's company-wide Identity Provider portal from scratch (React, Node.js, Terraform, Java); it became the default onboarding surface for every new service needing access control.
  • Built a refreshable cache in Go over Oracle Cloud object storage, cutting P99 latency and killing redundant API calls during burst traffic.
  • Built synthetic canary monitoring that continuously exercised the live auth flow (JWT → cert → regional login) against government-grade RBAC services, paging on threshold breaches; it caught region-wide outages and bad releases early, supporting 99.99% availability.
GoReact / NodeJavaTerraformRBAC99.99%
TRAJECTORY · SENIOR ENGINEER LEADlearned: the highest-leverage code review is a good 1:1.

Mission 04 · 2024–2025 · New York

Planet OptoLead Software Engineer · Opto Investments

An ice giant with a $300M ring system. First command post: leading a squad, aligning stakeholders, and making private markets legible.

  • Led a 4-engineer team building a fund-tracking engine over $300M+ in live assets, with cross-fund roll-up calculations, third-party integrations, and stakeholder-facing data viz.
  • Engineered event-driven transaction pipelines and real-time dashboards that ingest, process, and reconcile highly concurrent portfolio state changes with strict auditability and zero data loss.
  • Cut AI diligence latency 7s → <3.5s by refactoring the enrichment pipeline across every endpoint.
  • Prototyped and deployed internal APIs letting AI agents read structured investment data and take programmable actions across custom-fund workflows.
Team leadFintechAI agentsFastAPI
TRAJECTORY · LEAD FOUNDERlearned: there is no one to hand off to. the roadmap, the code, the invoice: all yours.

Mission 05 · 2025–present · The Frontier

Planet CodePoetFounder & Principal Engineer · you are here 📍

AI writes your code now. So who checks it against what you actually asked for? CodePoet turns intent into a contract, then hands you receipts every time the code drifts from it.

  • Built the trust loop: capture intent, approve it as a living contract, then a GitHub App that reviews every PR and flags behavioral regressions before merge. Running live across 50+ repos.
  • Engineered Tolkien, a deterministic tree-sitter engine that reads a codebase's real structure and standards, replacing flaky LLM scanning: 2s vs. 30–60s at $0 marginal cost, with call-graph blast-radius and drift scoring across 300+ symbol types.
  • Built a model-routing harness inside the Claude Code wrapper that sends each task to the right model, using eval infrastructure to pick the cheapest model that still passes, stretching a single subscription much further.
Intent → receiptsGitHub AppTolkienModel routingClaude + MCP0→1
TRAJECTORY · FOUNDER WHAT'S NEXTthe map ends here. that's the fun part.

Mission 06 · plotting a course

What's NextCoordinates still being set

CodePoet keeps shipping. Alongside it, I'm looking for the next place to point nine years of range: applied-AI work in fintech or developer tooling, as an engineering manager or technical PM, where I get to build the product and own the calls.

However it's framed, it gets built the same way everything in this system was: technology for the customer, starting from the customer. If that's the kind of person your team needs, the channel is open.

Systems check · Ship's instruments

What's on the console

Nine years of instruments, still all lit. Engineer depth, plus the lead and product roles that decide where the depth points.

Founder & Principal EngineerCodePoet · 0→1, solo
Lead Software EngineerOpto · 4-engineer team
Technical Product Managerproduct loops, GTM, evals

Languages

GoTypeScriptPythonJavaRustC++SQL

AI / LLM systems

Claude APIAgentic AIMCPTree-sitterRAGVector DBsLLM evals

Platform & infra

AWSTerraformDockerKubernetesPostgreSQLRedisCI/CD

Architecture

Distributed systemsEvent-drivenWebSocketsFastAPIReact / NodeObservability

Interlude · Strategy

The Knight's Gambit

Chess taught me the thing I use most at work: the best move is rarely the straight line. It also won me an Ohio state championship, so the knight and I go way back. My resume is in the vault. Guide the knight ♞ to the beacon ◆ to decrypt it.

🔒 VAULT STATUS: ENCRYPTED · awaiting knight at beacon square e5…

MOVES: 0 · PAR: 4

Interlude · Why "Curiosity"

They Called Us Curiosity

We spent centuries staring up and asking if anyone was out there. Then, instead of waiting for an answer, we built small brave robots. We gave them wheels, silicon brains, and names like Spirit, Opportunity, Perseverance, and Curiosity, and we sent them to other worlds to say: we existed, and we were the kind of species that wanted to meet you.

That's what stories are to me: signals we send so nobody has to feel alone. Every book on the shelf below is one of those signals. Honestly, so is this site.

"and they told us to tell you hello."from a post about the Mars rovers that I think about a lot
SpiritOpportunityCuriosity ← my callsignPerseverance
SOL 3,847 · MARS
UNIT: CURIOSITY · STATUS: STILL GOING

arrow keys work too. the rover likes company.

Interlude · Signals received

The Ship's Library

Every long flight needs a shelf. Not all of these are science fiction, but all of them are signals. They shaped how I think about systems, people, and the future. Pull one down.

Pick a spine ↑

the shelf is judging you politely

Click any book to see why it earned a permanent slot on a spaceship with strict weight limits.

Final transmission

Open a Channel

// coordinates: New York City
// status: CodePoet shipping · inbox open
// avg response latency: < 24 hrs(my APIs are faster, I'm working on it)

// external channels: GitHub · LinkedIn · CodePoet