---
title: "Neeraj Sharma (@neeraj0786) | Hashnode"
description: "Founder @ axionry.com | Head of Engineering @ AccioJob (YC'19). Read the latest articles by Neeraj Sharma on Hashnode."
url: "https://hashnode.com/@neeraj0786"
---

@neeraj0786

# Neeraj Sharma

@neeraj0786 · Gurugram, India · Joined October 2026

Founder @ axionry.com | Head of Engineering @ AccioJob (YC'19)

[Website](https://axionry.com/)

### [How a Voice Agent Stops Talking When You Interrupt: Designing Barge-In](https://axionry.hashnode.dev/how-a-voice-agent-stops-talking-when-you-interrupt-designing-barge-in?utm_source=hashnode&utm_medium=feed)

When Salesforce agreed to buy the voice AI company Cresta this month, one capability kept coming up in the coverage: its agents handle interruptions in real time. That sounds small. It is the feature

4 min read · [axionry.hashnode.dev](https://axionry.hashnode.dev/how-a-voice-agent-stops-talking-when-you-interrupt-designing-barge-in?utm_source=hashnode&utm_medium=feed)

[16h ago](https://hashnode.com/posts/how-a-voice-agent-stops-talking-when-you-interrupt-designing-barge-in/6afde5eb38333daf2fccdf41)

### [Uber Put 5,000 Agent Tools Behind One MCP Gateway. Here Is the Design, and a Small Version You Can Build.](https://axionry.hashnode.dev/uber-put-5-000-agent-tools-behind-one-mcp-gateway-here-is-the-design-and-a-small-version-you-can-build?utm_source=hashnode&utm_medium=feed)

Uber now runs more than 800 MCP servers and 5,000 agent tools behind one gateway. InfoQ covered the design on November 2, and it is the clearest public picture so far of what "agents in production" lo

5 min read · [axionry.hashnode.dev](https://axionry.hashnode.dev/uber-put-5-000-agent-tools-behind-one-mcp-gateway-here-is-the-design-and-a-small-version-you-can-build?utm_source=hashnode&utm_medium=feed)

[2d ago](https://hashnode.com/posts/uber-put-5-000-agent-tools-behind-one-mcp-gateway-here-is-the-design-and-a-small-version-you-can-build/6afc127854f2db3e25ac7e59)

### [System Design for a Private AI Video Studio: One GPU, Four Open Models and an Agent That Directs](https://axionry.hashnode.dev/system-design-for-a-private-ai-video-studio-one-gpu-four-open-models-and-an-agent-that-directs?utm_source=hashnode&utm_medium=feed)

A hosted AI video tool charges per credit and owns your workflow. A private studio on open models charges you per GPU minute and lets you keep your characters, prompts and outputs. We run one for our

5 min read · [axionry.hashnode.dev](https://axionry.hashnode.dev/system-design-for-a-private-ai-video-studio-one-gpu-four-open-models-and-an-agent-that-directs?utm_source=hashnode&utm_medium=feed)

[Nov 11](https://hashnode.com/posts/system-design-for-a-private-ai-video-studio-one-gpu-four-open-models-and-an-agent-that-directs/6af423fae410f1390d7b7b7a)

### [System Design for an AI Dubbing SaaS: One GPU Queue, a Consent Vault and Per-Minute Metering](https://axionry.hashnode.dev/system-design-for-an-ai-dubbing-saas-one-gpu-queue-a-consent-vault-and-per-minute-metering?utm_source=hashnode&utm_medium=feed)

How to design an ElevenLabs-style AI dubbing SaaS on open models: GPU job queue, consent vault, line-level edits and per-minute metering.

5 min read · [axionry.hashnode.dev](https://axionry.hashnode.dev/system-design-for-an-ai-dubbing-saas-one-gpu-queue-a-consent-vault-and-per-minute-metering?utm_source=hashnode&utm_medium=feed)

[Nov 10](https://hashnode.com/posts/system-design-for-an-ai-dubbing-saas-one-gpu-queue-a-consent-vault-and-per-minute-metering/6af2c3cf90fd0b0d5cfd6622)

### [System Design for a Daily AI Animated Series: One GPU Queue, a QA Gate and Three Output Formats](https://axionry.hashnode.dev/system-design-for-a-daily-ai-animated-series-one-gpu-queue-a-qa-gate-and-three-output-formats?utm_source=hashnode&utm_medium=feed)

We publish an animated nursery rhyme every day on Hummingtree: a 2 to 4 minute YouTube episode, a YouTube Short and an Instagram Reel. Our second episode needed 63 generated shots, plus original songs

6 min read · [axionry.hashnode.dev](https://axionry.hashnode.dev/system-design-for-a-daily-ai-animated-series-one-gpu-queue-a-qa-gate-and-three-output-formats?utm_source=hashnode&utm_medium=feed)

[Nov 5](https://hashnode.com/posts/system-design-for-a-daily-ai-animated-series-one-gpu-queue-a-qa-gate-and-three-output-formats/6aec6a0fd002957e975e57c5)
