Hi, I am Aniruddha Adak. I build AI agents from Kolkata, India, and this post is the single place where I put everything about what I do, what I have shipped, and where you can find me. If you have seen one of my articles before, you know I like to keep things simple and honest. So here is the full picture, no fluff, just the real work.
Who I am
I am a self taught style builder. By day I am a Computer Science student at Budge Budge Institute of Technology in Kolkata, class of 2026. By night, and honestly also by day, I write code, test every new AI model that ships, and send pull requests to open source projects I use and respect.
People ask what an AI Agent Engineer actually does. In plain words, I build software that can plan steps, make choices, and finish tasks on its own, with very little hand holding. I also work as a full stack developer, because an agent is only useful when it lives inside a real product that real people can touch.
My pattern has always been the same. Start big, fix later, keep shipping. It sounds chaotic, and sometimes it is, but it has taken me further than any careful plan ever did.
My skills, laid out properly
I keep my toolbox in three pillars. Modern JavaScript for products, Python for intelligence, and a long list of AI tools that I test every single day.
Frontend and product engineering
Area What I work with Frameworks Next.js, React.js, Astro, Qwik, SolidJS, Vite Languages TypeScript, JavaScript, HTML5, CSS3, some Kotlin for Android Styling and motion Tailwind CSS, Bootstrap, shadcn/ui, Framer Motion State and data Redux Toolkit, Chart.js, D3.js, WebSocketsBackend, data, and infrastructure
Area What I work with Server side Node.js, Express.js, FastAPI, REST APIs, GraphQL Databases MongoDB, PostgreSQL, MySQL, Firebase, Neon, SQLite, LanceDB, Prisma Cloud and DevOps Google Cloud, AWS basics, Docker, CI/CD, Vercel, Netlify Core languages Python, Java, C++, C, SQL Testing Jest, Mocha, VitestAI, machine learning, and agents
Area What I use or do ML foundations Python, TensorFlow, NumPy, neural networks Agent frameworks LangChain, LangGraph, CrewAI, multi agent orchestration Applied AI RAG pipelines, LLM integration, prompt engineering, speech to text with Whisper and AssemblyAI Daily AI toolbox Gemini, Claude, ChatGPT, Grok, DeepSeek, Qwen, Kimi, Mistral, Ollama, Perplexity, Antigravity, Hugging Face, GitHub CopilotMy GitHub, by the numbers
I work through two active GitHub accounts, plus a small side one. These numbers come straight from GitHub search on the day I wrote this, and they move up almost daily.
Account What lives there Scale aniruddhaadak80 Main account. AI agent projects, portfolios, open source work 126 public repos, 364 plus merged PRs, 689 plus issues opened AniruddhaAdak Earlier account. EchoCraft, VocalScribe, CodeMorpher and friends 18 public repos, 34 plus merged PRs aniruddhaadak9 Side account for experiments Newer, low activityOne thing I want to say about the issue count. Opening hundreds of well structured bug reports is a skill of its own. Each of my reports names the affected files, explains the risk in plain words, and proposes a concrete fix. Maintainers have labelled several of them as high quality, and a few were closed as completed within a day. That feels as good as a merged PR.
Open source contributions I am proud of
Anyone can fork a repo. Getting a fix reviewed, approved, and merged into someone else’s serious project is the real test. Here are merged pull requests I am genuinely proud of, newest first, every one of them verifiable.
Project What I changed Merged PR openclaw/openclaw Fixed fetchWithTimeout silently dropping the caller’s AbortSignal Jul 2026 #102951 openclaw/openclaw Replaced a broad Windows skip in browser tests with a dynamic symlink capability check Jul 2026 #90365 openclaw/openclaw Made install safe path symlink tests run on native Windows, 24 of 24 passing on a real Azure VM Jun 2026 #90275 openclaw/openclaw Made the QQ Bot symlinked media helper test robust on Windows Jun 2026 #90223 openclaw/openclaw Skills config wizard now shows an explicit all ready note when nothing needs installing May 2026 #85032 topoteretes/cognee Security fix: restricted global settings to superusers, masked API keys, added a registration switch Jun 2026 #3115 topoteretes/cognee Auto prefix Windows paths for LanceDB storage, fixing OS Error 3 on long paths Jun 2026 #3123 topoteretes/cognee Fixed visualize_graph() opening the wrong graph database Jun 2026 #3114 NousResearch/hermes-agent Hardened the permission bridge so a None response defaults to a safe deny Apr 2026 #13457 Tracer-Cloud/opensre Four merged PRs on docs and database root cause analysis logic Apr 2026 #924 vgvassilev/clad Documented the non_differentiable attribute with clear C++ examples Mar 2026 #1773 HSF/hsf.github.io GSoC 2026 website cleanups: metadata, terminology, template quoting fix May 2026 #1908 google-gemini/gemini-cli Removed the write todo list tool from UI tips Mar 2026 #22281 RocketChat/Rocket.Chat Docs fix, shipped in the 8.3.0 milestone Feb 2026 #38871 GetStream/Vision-Agents Corrected the Cartesia role description in the examples table Mar 2026 #366 sumanth-0/100LinesOfPythonCode Eight merged Python snippets during Hacktoberfest 2025 Oct 2025 #834Read that list and you will see my signature. I do not chase easy typo fixes. I go after Windows portability bugs nobody wants to touch, security holes in settings and permissions, and agent reliability issues. Then I prove the fix with real test output before asking for a review.
Bugs I found and issues I opened
My most recent work is a steady audit of OpenClaw through July 2026, covering security, memory, and performance. A sample of the freshest reports, verbatim from the tracker.
Projects I built myself
My repos read like a tour of everything I am curious about. Voice tools, AI resumes, agent marketplaces, games, dashboards, and more portfolios than anyone strictly needs.
Project What it does Stack smart-resume-analyzer AI resume analyzer and interview coach with ATS scoring and mock interviews Next.js, Gemini, Prisma agentgithub An agent native, GitHub like code forge with real git backed workflows Next.js, Clerk, Neon agentforge A marketplace where AI agents discover, hire, and pay each other in USDC Next.js, web3, Gemini omni-dev An autonomous agent swarm wired into more than 35 tools TypeScript, Gemini SecureOps-AI Multi agent DevSecOps pipeline that scans, enforces, and auto remediates JavaScript, MCP ai-agents-duel A narrative AI agent command center with live dueling Next.js, sketchbook UI herclew Self improving reasoning agents plus messaging gateways Python, TypeScript EchoCraft Turns audio into text, translations, and blog posts React, AssemblyAI, Gemini VocalScribe Fast speech to text for clean writing React, Node.js studystream AI learning assistant with quizzes and achievements Next.js, Algolia eventmatch Conversational AI event discovery Next.js, Algolia Lumina-CLI Turns a plain README into a cinematic 3D landing page JavaScript, CLI CodeMorpher Converts code between programming languages TypeScript my-android-portfolio My first Kotlin build, a portfolio app for Android KotlinAnd then there is the YouWare story. In early 2026 I gave myself a strange challenge and shipped 27 working AI applications in 45 days without writing traditional code. I wrote the whole experiment up as one of my most read articles. Love or hate no code, the output rate is hard to argue with.
My tweets that travelled
On X I run a personal AI news feed under @aniruddhadak. My bio says it straight: I test and share all the latest updates on AI, AI agents, and LLM releases. Since October 2024 I have published 54.8 thousand posts. That is well over a hundred posts a day, and yes, my thumbs are fine.
The posts people still bring up: the thread where I posted 49 AI generated images in one go and broke my own engagement record. A side by side of GPT Image and Gemini Nano Banana Pro drawing an infographic of how a coffee machine works. The time I asked Google Antigravity to build my perfect portfolio and it produced 56 sections. And a February 2026 post about computer use agents, where I said AI is moving from chatting to acting. I still believe that.
Articles I keep getting messages about
Title What it covers I’m Aniruddha Adak: The AI Agent Engineer from Kolkata, and here is my story My full self written profile I’m Aniruddha Adak, And This Is My YouWare Addiction: 27 Apps In 45 Days The no code sprint Hey World, I’m Aniruddha Adak, The Guy Who Posted 49 Images In One Thread The story behind my most famous thread Building an AI Powered Recipe Assistant with Agentic Postgres Agents meeting Postgres, a build log Contribution Chronicles: My Epic Hacktoberfest 2025 Adventure Field notes from a 200 plus PR HacktoberfestAchievements and certifications
Achievement Detail Hacktoberfest 2024 and 2025 More than 200 pull requests completed in each of two consecutive years, Holopin badges on my profile 27 apps in 45 days A self directed build sprint, fully written up on DEV Google Cloud Skills Boost Introduction to Large Language Models, Introduction to Generative AI Skillsoft Natural Language Processing and Deep Learning certificates LinkedIn AI Foundations certificate CodeStreet Hackathon 2023 First place, leading a team Academics 92 percent in Class XII, 83 percent in Class X, West Bengal BoardEverywhere you can find me
Where I am heading
My bet is simple. The next big shift in software is not more code, it is systems that reason, plan, and act with very little hand holding. Everything I do points the same direction. The agent marketplaces, the multi agent pipelines, the build sprints, the daily testing of every new model that ships. I am not waiting for the agent era to arrive. I am already building inside it, one pull request, one article, and one late night chai session at a time.
If you made it this far, thank you for reading. Say hi on X or Telegram, tell me what you are building, and if you maintain an open source project with a bug nobody wants to touch, you already know where to find me.
답글 남기기