Abhinav Singh
Summary
I build AI agents that do real work: assistants with tools, memory, and backends that run their jobs reliably. My projects pair each agent with its infrastructure, queues, APIs, deployment, from video studios to voice pipelines.
Now Building
An automated dubbing pipeline for Indian languages, built with Go and message queues.
Now Learning
Agent evaluation benchmarks, Kubernetes internals, real-time streaming audio, and AI infrastructure.
Skills
GoTypeScriptJavaScriptPythonC++SQLLLM orchestrationMCP / WebMCP toolingRAG + Vector DBAgent memory systemsSpeech & audio MLNode.js + ExpressFastAPIPostgreSQL & SQLRedis queues & background workersBrowser automation (Playwright)LinuxGitDocker & ComposeAWSKubernetes (learning)AI infra & evals (learning)Distributed systemsSystem design
Projects
Here are all of my favourite projects that I worked on
1. Panoptik: Agentic Video Studio
A video editing studio that runs in the browser, where an AI assistant suggests edits like zooms and subtitles for you to approve.
TypeScriptNext.jsWebMCPWebCodecs
2. SVAR: Hindi Call Analytics
A system that listens to Hindi customer calls and reports on agent behaviour, customer mood, rule violations, and overall call quality.
PythonFastAPIRedis RQPostgresVoice EngineeringReact
3. Signal: No-Code Web Monitor
A tool that watches any webpage and emails you when something changes, like a price dropping below the amount you set.
TypeScriptElectronPlaywrightBullMQLLM as a compiler
4. SunoHQ: Voice Agents
A voice assistant service for small businesses that talks to customers in Hindi and English over chat and voice messages.
PythonSTT/TTSRAGTelegram
5. Insight: AI News Feed Accumulator
A personal news feed that collects tech articles from many sites and sends you short AI-written summaries every day by email.
TypeScriptReactExpressSupabaseAI OrchestrationRAGPdf ParsingPlaywright
6. Lumio: Permaweb Moderator
AI community helper that watches chat groups, removes spam and scams, answers common questions, and never needs a break or sleep.
TypeScriptLuaRAGDiscord/Telegram botsAOWeb3Sentiment analysis
Open Source Experience
Apart from my own projects, I contribute to other codebases
CNCF AgentCube: 6 Merged PRs
Concurrency, garbage-collection and sandbox-lifecycle fixes in the agent session runtime.
Go
CNCF llm-d: 4 Merged PRs
Durable Redis dequeue via lease-claims, deadline and token metrics.
Go
better-auth: 1 Merged PR
Cross-tab session broadcast on sign-out and user update, with tests.
TypeScript
Pangolin: 2 Merged PRs
Session-table and cookie-domain fixes in the tunnelled proxy.
TypeScript
Hackathons
I have entered many hackathons so far. These are some of my results
1. Surreal World Asset Buildathon
Winner
Built Kreon Labs: decentralized IP provenance and infringement detection with C2PA, watermarking and Story Protocol.
Online1st place of 200+ teams
2. Agents of the Permaweb
Finalist
Built Lumio: a decentralised automation layer connecting Web2 APIs to the Permaweb.
OnlineTop 6 of 100 teams
3. World Computer Hacker League
Finalist
Built Fraktal: a Zapier-like automation platform for Web3 with Azle and ICP.
Asia-level selectionICP
4. PSB FinShield Hackathon
Finalist
Built Sentra: a graph-based fraud detection and adaptive risk-scoring engine for real-time authentication.
IIT HyderabadTop 50 of 2000+ teams
Education
KIET Group of Institutions
Contact
Hello,
If you have a role, a project, or just a question, I would be happy to hear from you. I usually reply within a day.
Abhinav Singh