Experience & highlighted work
Atlassian, General Motors and Indiana University, plus the Artificial intelligence — Software that performs tasks that normally need human judgment, such as language or image understanding.-native projects since.
Highlighted Work
Org Delete Capability
Led design and delivery of self-service org deletion, eliminating ~1K manual support tickets/year (5% of total volume). Architected a multi-phase async workflow — soft-delete, 30-day grace period, hard-delete — orchestrated across 12 downstream systems with full atomicity and audit compliance. Built with a team of three in under two quarters.
Integration Automation Framework
Standardized new Atlassian product onboarding onto the admin and identity platform. Every service in the chain used to need a code change to know a product existed; config-driven onboarding replaced that with a YAML Ain't Markup Language — A human-readable text format commonly used for configuration files. entry. End-to-end integration went from ~15 months to under a month, admin-side from 2–3 months to under 2 weeks, with ~5 products onboarded in ~6 months.
Adminhub Search
Feature lead for Adminhub Search serving ~1.One million — Shorthand in prices and limits, as in $1 per 1M tokens. customer organizations. Drove search quality improvements that boosted conversion from 15% → 35% and cut the no-results rate from 52% → 28% between beta and General availability — The stage when a product is released to everyone, after any preview or beta..
This Portfolio
An Artificial intelligence — Software that performs tasks that normally need human judgment, such as language or image understanding.-powered portfolio built with Next.js and FastAPI, fronted by a Claude agent grounded entirely in a private markdown knowledge base — with Langfuse tracing on every turn. Features live LeetCode stats, an interactive career graph, and a recruiter booking flow.
Homescreen
A personal work-and-life dashboard I built to replace my phone — one calm home screen instead of a dozen apps and a social feed. It runs on a module registry (Next.js 15, Drizzle, Neon Postgres): each tracker is a self-contained module that plugs into a shared shell, so it grows without a rewrite. Every integration sits behind a common connector interface, so the whole app can run on synthetic fixtures instead of live data; an interactive copy of its ranking, running on fixtures, is on this site's Artificial intelligence — Software that performs tasks that normally need human judgment, such as language or image understanding. Projects page.
Multi-Agent Orchestration Toolkit
A reusable multi-tier agent pipeline — spec → implement → review — applying distributed-systems rigor to Large language model — A neural network trained on large amounts of text to generate and understand language. agents: explicit contracts, verification gates, and failure-mode handling instead of ad-hoc prompting.
Experience
- —Built and shipped this portfolio (snehilv.dev) — a knowledge-base-grounded Claude agent with Langfuse tracing, prompt caching, and a capture loop that logs every question the Knowledge base — A collection of reference documents a system reads from to answer questions. could not answer; Next.js 15 + FastAPI
- —Built a personal dashboard (Homescreen) on a connector-interface architecture with live third-party integrations, magic-link auth, and on-demand refresh instead of polling; Next.js 15 + Neon PostgreSQL
- —Built a reusable multi-agent orchestration toolkit — a multi-tier spec → implement → review pipeline applying distributed-systems rigor to Large language model — A neural network trained on large amounts of text to generate and understand language. agents: explicit contracts, verification gates, failure-mode handling
- —Built a config-driven integration framework for product onboarding: end-to-end ~15 months → under a month, admin-side 2–3 months → under 2 weeks
- —Led Adminhub Search for ~1.One million — Shorthand in prices and limits, as in $1 per 1M tokens. customer orgs; improved search conversion 15% → 35%, cut no-results rate 52% → 28%
- —Led Org Delete — self-service org deletion fanned out across 12 downstream systems behind a 30-day grace window; 3 engineers, 6 months to General availability — The stage when a product is released to everyone, after any preview or beta.. Estimated ~1K support tickets/year removed
- —Mentored 5+ engineers through design reviews, architecture planning, and engineering best practices
- —Full-cycle owner and Site Reliability Engineering — The discipline of running software services reliably, using engineering practices. across Amazon Web Services — Amazon's cloud computing platform.-based services; founded and led the department security champion group (2021–2026), holding ~30–35 security postures above 90%
- —Built embedded Over-the-air (update) — Delivering software or firmware updates to a device wirelessly, with no physical connection. software and test infrastructure using C/C++ on QNX real-time operating system — A commercial real-time OS used in embedded and automotive systems./Linux for General Motors — A US automaker. Infotainment systems
- —Led C# integration test framework simulating Over-the-air (update) — Delivering software or firmware updates to a device wirelessly, with no physical connection. hardware, cutting 80% validation time and human hours
- —Implemented secure multi-party computation using Intel Intel Software Guard Extensions — A CPU feature that runs code in a hardware-protected enclave. and Obliv-C with side-channel attack protections
- —Engineered constant-time functions optimized down to Central processing unit — The main chip that executes a computer's instructions. clock precision
Ask me about Snehil
I know his work, background, and what he's looking for. Ask anything.
Responses are grounded in Snehil's knowledge base only.