Headline

Building AI Agents that matters | FDE @Push Protocol

About

Integrations Engineer by title, Forward Deployed Engineer by function I like building the parts of a system that most people never see: the connections, the stuff that has to work quietly so everything else can work loudly. There's something satisfying about watching a pipeline hold under real load after you've spent hours figuring out why it kept falling over. What I actually spend my time on: Scale. Not as a buzzword, as a personality trait at this point. High throughput, low latency, systems that need to survive real load, not just a nice architecture diagram. I get genuinely excited reading about consistent hashing, queues and distributed systems, which I recognize is not a normal thing to be excited about. Agentic AI, the real version. Not "wrap a prompt in a for loop and call it an agent." I'm talking actual agent architecture, agent security (an agent with tool access and zero guardrails is a chaos gremlin waiting to happen), and workflows that need to hold up outside of a demo where everything conveniently works on the first try. Cranberry Labs, my own thing, where I build solutions for businesses that are quietly drowning in manual work and losing cash and time. Think of it as me convincing companies to let systems do the boring parts so humans can do the parts that actually need a human. Basically: if it needs to scale, whether that's a system, a business, or apparently my own attention span across five things at once, I'm interested. Come argue with me about architecture decisions, I promise I have opinions and at least half of them are correct.

Location

Santuri

AI stack

claude code

gemini

langchain

langgraph

railway

Wispr Flow

Work experience

Software Engineer at Push Protocol | Previously EPNS, Sep 2023 - Present · 3 yrs 1 mo, Belize, 1. Designed a 30+ index strategy across 12+ collections, cutting hot-path queries to sub-100ms and sustaining 99.99% uptime for 80–100K users at 70 req/sec on a single-instance serverless deployment. 2. Built a low-latency, real-time transaction tracking backend for Push Chain DApps, handling 60,000+ requests/day with peak throughput of 1.3M requests/day. 3. Prototyped an intent-based cross-chain architecture, including a solver execution environment with priority mempool, order batching, custom relayer logic, and proof-generation for intent execution. 4. Owned partner integrations end-to-end, authoring onboarding docs and technical guides adopted by 10+ ecosystem partners including Uniswap, QuickSwap, Aave, and 1inch.

Education

Bachelor's degree, Computer Science, Indian Institute of Information Technology Vadodara

Higher secondary, Delhi Public School - India

Public links

github: https://github.com/Arunava-Barua

leetcode: https://leetcode.com/ArunavaBarua

arv-3d-portfolio.vercel.app: https://arv-3d-portfolio.vercel.app/