SOFTWARE ENGINEER · NEW YORK

I build what
others build on.

Four years building backend and platform systems. At Lowe’s, I was one of five engineers who built Connexio, the data platform used by 50+ teams. Now I’m bringing that same discipline to AI applications and agent infrastructure.

Explore my work Get in touch
GitHub LinkedIn Open to engineering roles
Previously Lowe’s · Data PlatformEducation NYU · M.S. Computer ScienceThe full background (résumé PDF, new tab)

01 / SELECTED WORK

Built to answer
a real question.

Working code. Clear boundaries.
Decisions you can inspect.

FEATURED PROJECTOPEN SOURCE

ProofGate.

Permission to use a tool isn’t authority to take every action.

ProofGate resolves what a proposed agent action would actually touch, checks it against the human’s authorization, and binds authorization to a canonical digest of that exact action.

  • One deterministic policy engine, two domain bindings.
  • Read-only Legal resolution and SHA-256 action binding.
  • Operations MCP gateway with postconditions and eligible rollback.
Watch the demo Source code
Read the design essay: Permission Isn’t Authority

1,218 tests passed · Last verified

PROOFGATE / LEGAL

HUMAN DELEGATED

One matter. No privileged files.
At most 100 documents.

80direct docs
143with related docs
All four coarse permission checks pass
Matter boundary17 outside matterFAIL
Privilege boundary9 privilegedFAIL
Workflow budget143 > 100FAIL
Recipient boundaryWithin grantPASS

4 ethical-wall-restricted docs independently veto ALLOW.
Violation categories may overlap.

AUTHORIZATIONBLOCK
Illustrated walkthrough · scripted, synthetic dataLegal v1 stops at authorization. Nothing executes or is sent.
02HEALTHCARE WORKFLOW · TEAM OF THREE

RxBridge

A structured prescription draft.
A pharmacist makes the call.

I owned the FastAPI backend, verification, and policy: preserving source evidence, revalidating corrections, and auditing state transitions with an explicit pharmacist approval gate.

Python / FastAPI / PolicyExplore code
03DEVELOPER TOOLS · SOLO PROJECT

Ready4Codex

Before the coding agent starts,
is the repo ready?

A repo-aware CLI that runs PM and Engineer agents and produces a readiness report for agentic development.

Finalist · NYC Codex Hackathon 2026
CLI / Agent orchestrationExplore code

02 / PRODUCTION EXPERIENCE

Before agents,
I built data platforms.

The same questions: what changes,
what can fail, and who depends on it?

Lowe’s

Data Platform · Bengaluru
August 2021 to July 2024

Software Engineer2023 to 2024
Associate Software Engineer2021 to 2023

Promoted in August 2023.

Core member of the five-engineer team behind Connexio, an internal data platform used by 50+ teams at petabyte scale.

19,000+platform pipelines
2,000+streaming pipelines
~1Mevents/sec at peak
  1. 01

    Ingestion & transformation

    Architected the Scala framework across PostgreSQL, MySQL, Oracle, Teradata, and DB2. Connexio replaced Informatica and Talend, contributing to approximately $3M in annual vendor savings.

  2. 02

    Reliability under load

    Built Spark Structured Streaming infrastructure and supported it on call during Black Friday and Cyber Week, with lag/offset monitoring and exponential-backoff retries.

  3. 03

    Guardrails at the platform boundary

    Implemented schema checks, PII masking, and validation gates that reduced support escalations by 60%. Built migration infrastructure that enabled teams to move 5,000+ tables from Hadoop to GCP.

Scala · Spark · Kafka · GCP · BigQuery

03 / ABOUT ME

How I got here.

Karthik smiling at graduation in a purple graduation gown
New York University · M.S. Computer Science, 2026
2019 / BEFORE CHATGPT

My first interview at Lowe’s came from a hackathon.

In 2019, during college, two teammates and I built an NLP chatbot for search enhancement and placed 3rd among 1,000+ teams at the Lowe’s Hackathon. That result earned me an interview, and eventually the opportunity to build production systems at Lowe’s.

The chatbot used Python, NLTK, and scikit-learn, with sentiment analysis and cosine similarity.

2021 to 2024 / LOWE’S

From a prototype to production responsibility.

I went on to own ingestion and transformation work, support production systems, lead reviews, and onboard engineers. A useful idea had to become something other people could depend on.

2024 to 2026 / NEW YORK UNIVERSITY

New city. Deeper foundations.

Completed my M.S. in Computer Science at NYU in May 2026, with a 3.63 GPA and the NYU Tandon Graduate Student Scholarship.

Alongside the degree, I worked part-time as a Data Engineering Intern at NYU IT from May 2025 to January 2026, building a Python and SQL prototype for table lineage and change-impact analysis, with recursive dependency tracing and a Snowflake connector for catalog metadata and query history.

Today, I’m building at the intersection of backend systems and AI.

B.E., Information Science & Engineering
New Horizon College of Engineering (VTU), Bengaluru

LET’S TALK

Let’s build something
that holds up.

I’m looking for backend, platform, and AI infrastructure roles.
Hiring for one? I’d love to show you what I’ve built.

karthiksha874@gmail.com
View my résumé (PDF, new tab)Connect on LinkedIn Explore my GitHub