unfair.so
Browse more creators
unfair.so/x/@vishalsingh2972
Vishal Singh
Vishal Singh12.4K
Vishal Singh
This creator added themselves. Not verified by unfair.so.

Applied AI Engineer • DevRel 🥑 I build Custom Voice AI & Agents for businesses DM for Work / Paid Collabs ☕ Shipping AI Products in Public @debugginglife25

@vishalsingh2972AI AgentsAI ToolsAI Apps
12.4KFollowers
4.4KMedian impr. / post
1.0%Engagement
22+17
Paid partnerships
In their audience
6 from @ycombinator follow Vishal Singh
@ishaansehgal@louis030195@daleverett@Totzenberger@rpatm@shijiew_4 affiliated with @apple3 affiliated with @github3 affiliated with @sarvamai3 affiliated with @microsoft2 affiliated with @twilio
849 verified followers analyzed
@ishaansehgal@louis030195@daleverett@Totzenberger@rpatm@shijiew_+84384 AI engineers67 engineers55 AI founders46 founders38 devtools engineers27 AI creators
25–34 core age group
40.6%
28.1% of audience in United States
🇺🇸
Oct 10 – Oct 1683 posts analyzed · replies not counted · updates weekly
ImpressionsEngagements
4.8K2.4K0
104520
Oct 10Oct 11Oct 12Oct 13Oct 14Oct 15Oct 16
Past collabs
Vishal Singh@vishalsingh2972 · Oct 21WHAT THE ACTUAL FUCK?! Anthropic just decided to go after OpenAI Dots and Grok Bot with Claude Duet, powered by the absolute MONSTER that is Opus 5.5. We literally go to sleep, wake up, and somehow Anthropic has dropped ANOTHER product to take on its competitors. LIKE HOW ARE THESE GUYS SHIPPING THIS FAST?! THEY JUST CAN'T STOP WINNING. What an absolutely LEGENDARY run. 😭ClaudePaid partnership566K views
Vishal Singh@vishalsingh2972 · Oct 20Open source vs closed source. Light vs darkness. Two very different futures for humanity. This is why I love DeepSeek. Free research. Open models. And now, even their production inference stack. Community over everything. Can't imagine OpenAI, Anthropic, Google, or Meta doing this.DeepseekPaid partnership87.2K views
Vishal Singh@vishalsingh2972 · Oct 15I built a local open-source version for GPT-Live-1 earlier. Waiting on the API to do the same for 1.1 GPT-Live running fully on your machine: 😌 Silero VAD → Faster-Whisper → local LLM → Supertonic TTS Handles barge-in, playback-aware history, and pauses properly. Still a POC but fun to run fully local. Demo + GitHub repo in the commentsOpenaiPaid partnership5.7K views
Vishal Singh@vishalsingh2972 · Oct 12I built an AI tool that lets you change what you said in a video just by typing new words. 🤯 And the crazy part? It recreates your voice AND syncs your lip movements to match the new words. Same video. Different words. Zero re-recording. 💀 I was tired of re-recording the same demo for every new hackathon. Sometimes, literally the ONLY thing that changed was the hackathon name 😭 I thought... why can't editing what I said be as easy as editing text? So, I built 𝗦𝗽𝗲𝗲𝗰𝗵 𝗦𝘂𝗿𝗴𝗲𝗼𝗻 🔥 Here's how it works: → Edit the words in your transcript → AI generates the new speech in your own voice → AI syncs your lip movements to match Updated video in minutes. No retakes. No manual video editing. Imagine fixing a YouTube mistake, updating an old tutorial, or changing your startup demo without recording everything again. Built with @FishAudio 🦈 Here's a quick before-and-after demo 👇 See if you can spot the edit 😉FishPaid partnership3.4K views
Vishal Singh@vishalsingh2972 · Sep 23wanted to test Gemini 3.8 Flash + I hate making Google Forms manually so built baatsheet 😭 ~ talk to it, get a production-ready form + AI response insights in seconds fun little voice ai project Quick demo + GitHub repo in the comments 👇GooglePaid partnership4.5K views
Vishal Singh@vishalsingh2972 · Sep 18“What if @elonmusk was born in Bihar?” Made this AI audio episode in 2 min 😂 Sherpa turned it into a full series with Trump as the villain, Sam Altman + Dario Amodei as the jealous cousins, and Elon's journey from Bihar to Mars 🚀 possibilities especially for 🇮🇳 short storytelling is insane with this Quick 2 min demo 👇PocketfmPaid partnership1.6K views
Vishal Singh@vishalsingh2972 · Sep 18jev is hottest topic in tech rn 🔥 but how would you describe jev to someone non-technical? lemme explain it in the simplest way possible 🌹 jev is really good at answering multiple choice questions (and telling you how confident it is) and a surprising number of real-world problems can be broken down into lots of small multiple choice questions, solved one after another so software can just act on the answers. that’s basically the main idea, awesome work @typesafeai 👏TypesafePaid partnership2.1K views
Vishal Singh@vishalsingh2972 · Sep 16Tested Gemini 3.8 Live on 3 of my favorite books and the results were insane 🔥 I pointed the camera at: 📖 Bhagavad Gita — asked it to explain a Sanskrit sloka + make it relevant to a 20s Gen Z developer 🧠 Atomic Habits — showed it a diagram and asked for a quick explanation 😴 Chicken Soup for the Soul — gave it a page and asked for a warm bedtime-story version No PDFs. No uploads. Just the phone camera + my ugly voice 😂 I genuinely didn’t expect it to understand the context this well. Awesome job @GoogleAI Team ❤️ Demo 👇GooglePaid partnership10K views
Vishal Singh@vishalsingh2972 · Sep 14Built a Voice AI agent in 5 mins, this is illegally good. Love it 😻 Opened CallKaro just to see how fast I could ship something real. Made a used-car buyer qualification agent for @myspinny 🚗: prompt → conversation flow → Hinglish → budget/intent extraction → guardrails → post-call summary Then I tried to break it like an actual buyer: “six fifty” “automatic better hai” “7 lakh ke andar milega?” “discount kitna doge?” “loan approve ho jayega na?” - It caught the vague answers. - Asked the right follow-ups. - Didn’t hallucinate discounts, availability, or loan approval. ✅ That’s where it gets super interesting. Making an agent talk is easy. But getting conversation logic + guardrails + structured extraction + natural voice to hold together in one call is the actual job. Pronunciation still needs some work, 5 mins was not enough for me to teach it everything lol 😭 Zero → a working conversational voice agent ready to deploy in ~5 minutes. 🤯 Honestly, crazy how fast this stuff is getting.CallkaroPaid partnership1.8K views
Vishal Singh@vishalsingh2972 · Sep 13Quick weekend Voice AI project 🔥 Built Jaldi Bol + tested it with a real visa processing/travel-tech company @atlys 📊 My 𝗟𝗮𝘁𝗲𝗻𝗰𝘆 𝗕𝗿𝗲𝗮𝗸𝗱𝗼𝘄𝗻: 🗣️ Deepgram Nova-3 → ~125ms 🧠 Gemini Flash Lite → ~180ms TTFT 🔊 Cartesia Sonic 3.6 → ~105ms TTFA 🎙️ Silero VAD → ~180ms ⚡ LiveKit WebRTC → ~50ms Put it all together → I got ~520ms p50 end-to-end latency (which I felt was pretty great ngl 🧘) I know how annoying it is when you don’t get a reply back, especially when you’re waiting for an answer and the clock is ticking. 😭 So, I made Jaldi Bol to quickly handle: → Visa status → Document issues → Urgent cases → Human escalation, if needed The flow here was pretty simple: Visa delayed → Jaldi Bol checks → answers → escalates if needed. Next, went a little deep on the voice loop part: ✅ Full-Duplex Barge-in: Instant interruption cutoff (<300ms) with zero echo feedback. ✅ Multi-Agent Orchestration: TriageAgent → VisaAgent → EscalationAgent ✅ Zero Hallucination: 100% grounded against visa database schemas. ✅ Short 1–2 sentence responses for natural conversations Biggest learnings (after aggressive testing): → Streaming everything matters more than just a faster LLM. → Barge-in is huge. Interrupt it → it shuts up and listens. → Shorter responses sounded way more human. Started this as a latency test. Wanted to see firsthand how fast I could push a voice agent before it actually started feeling conversational. Demo + GitHub repo in the comments 👇DeepgramPaid partnership1.6K views
Vishal Singh@vishalsingh2972 · Sep 5H1B visa interviews are scary af 😭 and I spend half my time in traffic anyways why not use that time practicing visa grilling lol so, build Visa Babu 🇮🇳 basically, a voice AI visa interviewer (but very real) it starts off nice then slowly gets more annoying depending on how you answer ~ filler words, pauses, pacing, confidence etc. finally, you get a visa verdict: Approved / Rejected with improvement tips Tech stack I used: 🧠 Sarvam AI → Bulbul v3 for Indic TTS + 105B for evaluation 🗣️ Cartesia → English TTS with British officer voice 📲 WebSockets → real-time voice wanted this to be something I can actually use while sitting in traffic, so made it mobile-first and tested it on 3G/4G speeds (worked like butter 🤤) then obviously I had to try breaking it, so tried: → 3G throttling + 5% packet loss - still worked fine → PCM vs Opus - worked smoothly → WebSocket streaming - solid Demo video + GitHub repo in the comments 👇CartesiaPaid partnership4K views
Vishal Singh@vishalsingh2972 · Aug 15Built JobJugaadu AI using @heycalle 🙈 Independence Day 🇮🇳 weekend mini project. most local businesses are hiring but never post jobs online. So, I made an AI that literally calls approved employers, asks "Bhai, are you hiring?", collects salary + shifts + openings, and turns the call into structured job data. Peak Indian jugaad, now with AI agents lol. repo link in the comments.Heycall-ePaid partnership2.4K views
Vishal Singh@vishalsingh2972 · Aug 13Sarvam AI opened Voice Agents today, so I built a Hyderabadi sabzi aunty 😭 Called her and bargained in Hindi + Telugu + English, kept interrupting, switched languages mid-sentence, and she still remembered my order, gave me a final total, and even confirmed my (fake) UPI payment. Took me ~15 mins to build. It was fun to recreate the lost art of vegetable bargaining, but this time with an AI agent.SarvamPaid partnership180K views
Vishal Singh@vishalsingh2972 · Aug 11okay Grok Bot is actually kind of ridiculous it can: - open a computer by itself - log into your apps - use websites like a human - work while you’re asleep - clean your inbox - send emails - update your CRM - research people and companies - write LinkedIn + email drafts in your style - remember how you work - learn a workflow after watching you once - run that workflow forever - coordinate with other bots - only bother you when it needs a decision .....and so much more that I'm yet to decipher 😋 this is less “AI assistant” and more “the intern who somehow became COO overnight lol” @spacexai team cooked 🙌XPaid partnership146K views
Vishal Singh@vishalsingh2972 · Aug 11Let’s understand @Graphify in less than 2 minutes. (Explained by someone who literally can’t go anywhere without Google Maps 😝)GraphifyPaid partnership1.9K views
Vishal Singh@vishalsingh2972 · Aug 8Wanted to test Bolna’s new Web Call SDK, so I built an Aadhaar-style address update using voice. 🇮🇳 Instead of manually typing, I simply spoke in Telugu, English, or even a mix of both, and the form fills automatically in real time. Bolna’s Web Call SDK + function calling + webhooks let the voice agent actually update details in the browser while I speak. GitHub link in comments. Thanks to @tanishasharmax for the quick beta access. Demo video👇BolnaPaid partnership3K views
Vishal Singh@vishalsingh2972 · Jul 28Some good news, Cursor is launching a exclusive $7 plan (Rs 649) just for 🇮🇳 More of this please @elonmusk , better AI accessibility for developers in India.CursorPaid partnership14.3K views
Vishal Singh@vishalsingh2972 · Jul 21Watching how LLMs think is way more fun than just reading its output. So, I built an interactive 3D visual debugger for transformer AI models. Tested with the latest @Alibaba_Qwen 3.8 model. • Live token generation • Tensor inspection • Architecture explorer • Real inference visualization Open Sourcing soon ❤️ Demo video 👇QwenPaid partnership2.8K views
Vishal Singh@vishalsingh2972 · Jul 12Weekend AI Project 🔥 Meet Research Saathi - a multi-agent AI research team that actually works. Instead of fighting one massive prompt, I orchestrated 4 specialized agents using LangGraph: 🔍 Search: Finds credible, recent sources. 📄 Reader: Extracts signal from the noise. ✍️ Writer: Synthesizes a structured report. 🧐 Critic: Reviews and kills hallucinations. The Workflow: The agents pass state back and forth in a loop, mimicking a real research firm. The quality jump compared to a single-LLM approach is night and day. Tech Stack: • LangGraph (Orchestration) • LangChain (Logic) • Streamlit (Interface) Key Takeaway: Agentic workflows > Clever prompting. Giving models clear roles and a feedback loop is the ultimate unlock for reliability. Demo video + GitHub in the comments. 👇LangchainPaid partnership3.5K views
Vishal Singh@vishalsingh2972 · Jul 11Built a real-time Voice Agent in 30 minutes using Sarvam's new MCP server. My mission? Solve the most important problem in Bengaluru: finding a good Masala Dosa 🥘 I wanted to test Sarvam’s localized models in a live environment. I don't speak Kannada, so I built and tested it in Telugu. It was wild how smoothly it handled language switching, real-time tool calling, and multi-turn conversations without noticeable lag. Tech stack 🔥 • @SarvamAI (saaras + bulbul) • LiveKit Agents • Cline Built in Telugu, but since Sarvam supports 22+ Indian languages, feel free to fork the repo and try it in your own language! GitHub repo link in the comments. Demo video 👇LivekitPaid partnership4K views
Vishal Singh@vishalsingh2972 · Jun 18Mini Project for Learning AI Agents: Modiji said, "Don't buy gold." Mom said, "Okay... but what's the gold price today?" 🤣 That conversation inspired my first AI agent. I'm learning AI agents, and the most fun way to learn is by building. So I built a simple agent that: ✅ Tracks gold prices automatically ✅ Detects price drops ✅ Sends my mom a Hindi WhatsApp voice note when gold gets cheaper Tech stack I used: 🧠 Brain: Gemini (via Vercel AI SDK) 🗣️ Voice: @SarvamAI Bulbul V3 📲 Delivery: Twilio + Cloudinary (for native WhatsApp audio) Sharing a quick demo video below 👇 GitHub repo link in the comments.TwilioPaid partnership6.1K views
Vishal Singh@vishalsingh2972 · Jun 13Super powerful local LLM just dropped 🔥 Kimi K2.7-Code is sick 👀 Opus 4.8/GPT-5.5 costs $20/m This one is better + free + self-hosted. Here's what you can do with Kimi K2.7-Code 👇: → Run it in your terminal as a full coding agent → Spawn 100+ sub-agents for parallel refactors → Multi-day autonomous coding sessions with persistent memory → Works inside VS Code, Cursor, or Zed → Connect it to Claude Code or Cline via API Available now on Kimi API and Kimi Code CLI. 6x High-Speed Mode coming soon. Go try it → http://kimi.com/codeKimiPaid partnership4.1K views
Planning a launch?Vishal Singh is one of 200 curated B2B creators. Tell us what you are launching and we will put the shortlist together.
[email protected]·@notmiramurati
Built with unfair.so
1 picked$120