Rohan Sharma
Portrait of Rohan Sharma

About Rohan Sharma

I am a voice AI engineer at Nugget by Zomato, in Delhi NCR.

I design real-time systems that pair low-latency, AI-powered automation with microservice architectures: streaming audio in, deciding when a person has finished speaking, reasoning about what to say, and answering inside a strict latency budget. Before voice I built agent hubs, an interview platform and search infrastructure at GoodSpace AI, and I still write Go libraries and MCP servers on the side.

Crowds gather around the Voder, an early electronic speech synthesizer, at the 1939 New York World’s Fair.
The Voder, an early electronic speech synthesizer, at the 1939 New York World’s Fair. Bell Telephone Quarterly, 1940. No known copyright restrictions, via Wikimedia Commons.

Experience

Nugget by Zomato

SDE-2, January 2026 to presentSDE, backend and AI agents, April 2025 to January 2026Gurugram, India

  • Built and own the core voice AI agent platform: a distributed, microservices-based system with low-latency, end-to-end orchestration for real-time conversations.
  • Work on the voice processing pipeline: voice activity detection, noise cancellation, audio normalisation and turn-boundary detection.
  • Architected the streaming systems that handle live audio ingestion, agent reasoning and response synthesis inside strict latency budgets.
  • Designed the orchestration layer for agent coordination, state, concurrency and fault tolerance.
  • Tune latency, throughput and cost for high-volume traffic.
  • Built and extended AI-powered social customer support and outreach across Twitter/X, Google My Business and Meta, with multi-channel ingestion, routing and human-in-the-loop escalation.
  • Promoted to SDE-2 within a year for end-to-end ownership.

GoodSpace AI

Tech lead, December 2024 to April 2025SDE, fullstack, June 2024 to December 2024SDE intern, January 2024 to June 2024Noida and Delhi, India

  • Launched an AI sales agent hub that managed more than 1,000 concurrent calls at under 300 ms on self-hosted call infrastructure, across voice, WhatsApp, email and SMS. Lead conversion rose 30%.
  • Engineered a low-latency multilingual AI interview platform: more than 25,000 interviews in over 55 languages, with proctoring, transcription and behavioural analysis.
  • Built a custom KNN vector search engine that ingested more than 12 million records at 10,000 documents a second and retrieved in under 100 ms.
  • Cut resume parsing from 12 seconds to about 3 with a custom NLP pipeline on NVIDIA GPUs.
  • Implemented a Kafka event-driven microservice that processed more than 500,000 events a day at under 10 ms.
  • Ran 15+ Dockerised microservices on 8+ VMs across AWS and Azure at 99.9% uptime.
  • Led a team of six and cut CI/CD build times from 8 minutes to about 1.5.

Sampark Softwares

SDE intern, May 2023 to August 2023Gurugram, India

  • Built a sentiment analysis model for a support chatbot with NLP and Node.js, with intent and context analysis, deployed as a REST API.
  • Built the speech recognition and text-to-speech interface in React.

Delhi Technological University

Web development intern (USIP), August 2022 to December 2022Delhi, India

Coding Ninjas

Teaching assistant, December 2021 to April 2022Delhi, India

  • Resolved and evaluated more than 600 doubts and projects in data structures, algorithms and object-oriented programming in Java, and mentored 50+ students.

Education

Delhi Technological University, B.Tech, production and computer engineering, 2020 to 2024.

Community

Certifications and awards

Tools

Go, Python, TypeScript, Node.js, Java, C++, FastAPI, gRPC, Kafka, RabbitMQ, Redis, PostgreSQL, MongoDB, OpenSearch, Docker, Kubernetes, Kuma, Envoy, Nginx, AWS, Google Cloud and Azure. On the AI side: voice activity detection, LLMs, RAG, embeddings, vector search, AI agents and the Model Context Protocol.

Say hello

Email is best: 2111rohan@gmail.com. I am also on LinkedIn, GitHub and Medium.