{"version":1,"url":"https://www.parasmadan.in/api/portfolio","profile":{"name":"Paras Madan","description":"I build AI products and help them find their market.","roles":["AI engineer","Founder","GTM operator"],"organization":{"name":"Varnan","url":"https://varnan.tech"},"sameAs":["https://www.linkedin.com/in/paras-madan-a9863716b/","https://www.instagram.com/parasmadan.in/","https://www.youtube.com/@parasmadan_in","https://x.com/ParasMadan9","https://medium.com/@parasmadan.in"],"url":"https://www.parasmadan.in/"},"services":[{"name":"AI product engineering","description":"Building production AI products and software."},{"name":"Go-to-market and developer distribution","description":"Helping AI and developer products reach their market and grow adoption."},{"name":"Technical content and collaborations","description":"AI and developer content, product videos, and brand collaborations."},{"name":"Applied AI workshops and speaking","description":"Practical AI sessions for enterprise teams, business leaders, and engineers."}],"portfolio":{"story":{"title":"I began with code.","subtitle":"The companies, experiments, and people that shaped my work.","introduction":["I started building things seriously around 2019. I was studying computer science and spending most of my time around machine learning, deep learning, research papers, hackathons, and small products I thought were worth trying.","AI was a very different space then. Most of my exposure came through models, datasets, multilingual systems, and figuring out why something that worked in a paper would completely fall apart when you tried to build it."],"chapters":[{"id":"iit-roorkee","title":"My first introduction to AI","paragraphs":["My first introduction to machine learning came during a research internship in the Electronics Department at IIT Roorkee. My professor gave me a task: use basic machine learning algorithms to predict MOSFET and capacitor values. That was where my interest in AI began.","The work made me curious about what else these models could do, so I started learning more about the field and experimenting on my own. I spent time with transformer-based language models such as T5, trying to understand how they worked and what I could build with them.","It was difficult to get things working. The amount of text I could pass to a model was limited, and I had to work directly with model code, tokenizers, and the Transformers library. Those early experiments were how I started learning to build with AI."]},{"id":"cruxe","title":"Cruxe","paragraphs":["Cruxe was my first proper attempt at building a company, and the idea changed more times than I can remember.","One version was something like Inshorts for interesting content. Another was a summarisation API for media companies. I somehow got six editors from The Times of India into a meeting while I was still in college, and they explained quite clearly why they would never buy it.","So we changed it again.","Cruxe eventually became a content recommendation product, something I used to describe as “Spotify for content.” We built a team around it and, at one point, reached around 100,000 users within a few months.","Running it taught me as much about finding users as building software. I had to write emails, explain the product, make content, talk to users, change the positioning, and understand why people came back or disappeared.","Eventually, Cruxe stopped. There were multiple reasons, including my health at the time, and I reached a point where I had to sit down and start preparing for software engineering interviews."]},{"id":"paytm","title":"Paytm and the internet","paragraphs":["I joined Paytm as a backend engineer and worked around merchant onboarding. It gave me a very different understanding of engineering because the systems were operating at a scale where small decisions could affect a lot of people.","At the same time, I kept posting online. I had started creating content during Cruxe because I needed people to discover what we were building, and I simply continued doing it.","Most of the early content did nothing. Over time, AI became the subject I kept coming back to because it was the area I understood best. Then generative AI started moving very quickly, and suddenly a subject I had been following for years had everyone's attention.","The audience grew to hundreds of thousands of people across platforms. More importantly, publishing constantly taught me how people react to technology. You can explain the same product in five different ways and immediately see which explanation people understand and which one they ignore."]},{"id":"ai","title":"Moving deeper into AI","paragraphs":["After Paytm, I moved deeper into AI companies.","At Fluid AI, I spent more time around enterprise AI and customers, including representing the company in rooms like Mastercard in Dubai. I realised I enjoyed understanding the product technically and then sitting with someone who needed to understand why it mattered.","Later, I worked with Unify AI and Athina AI, both YC W23 companies. My work moved across DevRel, community, onboarding, LLM evaluation, technical content, and distribution.","Because these were small AI teams, there was never a clean line between product and GTM. I was constantly moving between understanding what was being built and figuring out how developers or customers should discover it.","Around the same period, I was also teaching a lot. I have now taken more than 100 live sessions and workshops around AI, spoken at companies and institutions such as Mastercard, HSBC and Inc42, and given two TEDx talks.","Teaching made one thing very clear to me: you can understand a technology very deeply and still be terrible at explaining it to the person sitting in front of you."]},{"id":"varnan","title":"Varnan","paragraphs":["Varnan came out of everything that happened before it.","I kept meeting technical founders who had built genuinely good products but struggled to explain them, distribute them, or reach the right users. I understood the problem because I had been on both sides of it.","At Varnan, we worked with more than 15 AI and developer-product companies. The team grew to seven people, and we spent most of our time experimenting with positioning, technical content, launches, Reddit, communities, founder-led distribution, and whatever else could help a good product reach the right people.","Some things worked extremely well. Others did absolutely nothing. That experience has made me much more skeptical of marketing playbooks and much more interested in understanding the product, the market, and the people before deciding what distribution should look like."]},{"id":"now","title":"Where I am now","paragraphs":["Looking back, the path has moved through engineering, startups, content, DevRel, teaching, GTM, and building companies. It looks slightly random when written as job titles, but it has mostly been the same curiosity showing up in different forms.","I like understanding new technology, building around it, and figuring out how it changes the way people and companies work. That is still where I spend most of my time today."]}],"url":"https://www.parasmadan.in/story","context":"First-person account authored by Paras Madan."},"collaborations":{"url":"https://www.parasmadan.in/collaborations","context":"Portfolio-reported outcomes; supporting case studies and videos are linked per entry. Collaboration does not imply employment or endorsement.","entries":[{"name":"Pine Labs","focus":"Video Collaboration","impact":"52K views · 1,500+ interactions · 1,330+ likes","links":[{"label":"View post","href":"https://www.instagram.com/p/DdlumHpC0QH/"}]},{"name":"Tencent","focus":"Content Collaborations","links":[{"label":"Instagram post 1","href":"https://www.instagram.com/p/DdvhNgklFbh/?img_index=1"},{"label":"Instagram post 2","href":"https://www.instagram.com/p/Dcq0g3PiTIF/"}]},{"name":"Comnyang","focus":"Desktop Companion","impact":"3.3M video views · 524 orders","caseStudy":"https://varnan.tech/case-studies/how-varnan-scaled-comnyang-from-0-to-500-sales-in-a-week","video":"https://www.instagram.com/p/DZNGt0qCH2d/"},{"name":"OpenDirectory","focus":"Developer Package Directory","impact":"321K video views · 594 GitHub stars · 30K–40K npm downloads","caseStudy":"https://varnan.tech/case-studies/how-opendirectory-hit-10-000-npm-downloads-and-350-github-stars-in-one-month-from-scratch","video":"https://www.instagram.com/p/DX6qPpwCjtK/"},{"name":"LibrePods","focus":"Open-source Framework","impact":"490K+ video views · 5,500+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/8500-github-stars-across-7-dev-tools-varnans-predictable-distribution-engine","video":"https://www.instagram.com/p/DRY3nPaEhQM/"},{"name":"Git City","focus":"Open-source Tool","impact":"1M+ video views · 1,540+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/how-varnan-helped-git-city-reach-19-000-new-signups-in-48-hours","video":"https://www.instagram.com/p/DVbEJoUktVg/"},{"name":"PicoClaw","focus":"Open-source Project","impact":"450K+ video views · 1,500+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/8500-github-stars-across-7-dev-tools-varnans-predictable-distribution-engine","video":"https://www.instagram.com/p/DU3STPtkna5/"},{"name":"Server Survival","focus":"Open-source Tool","impact":"180K+ video views · 1,000+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/how-varnan-helped-server-survival-gained-1000-github-stars-and-built-an-active-contributor-base-in-90-days","video":"https://www.instagram.com/reels/DR6XS5TkuQm/b"},{"name":"Shannon","focus":"Open-source Project","impact":"390K+ video views · 500+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/8500-github-stars-across-7-dev-tools-varnans-predictable-distribution-engine","video":"https://www.instagram.com/p/DUfL99jkrKB/"},{"name":"Macless-Haystack","focus":"Open-source Project","impact":"450K+ video views · 450+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/8500-github-stars-across-7-dev-tools-varnans-predictable-distribution-engine","video":"https://www.instagram.com/p/DRws5zmkiU-/"},{"name":"AirLLM","focus":"Open-source Project","impact":"220K+ video views · 250+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/8500-github-stars-across-7-dev-tools-varnans-predictable-distribution-engine","video":"https://www.instagram.com/p/DVOfaRAkuj2/"},{"name":"Pixel Agents","focus":"Open-source Project","impact":"160K+ video views · 200+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/8500-github-stars-across-7-dev-tools-varnans-predictable-distribution-engine","video":"https://www.instagram.com/p/DVVSUT1ki9B/"},{"name":"CraftGPT","focus":"Open-source Tool","impact":"343K+ video views · 180+ GitHub stars","caseStudy":"https://varnan.tech/case-studies/8500-github-stars-across-7-dev-tools-varnans-predictable-distribution-engine","video":"https://www.instagram.com/p/DPTnLNWgTQo/"},{"name":"Android-MCP","focus":"Open-source Project","impact":"70K+ video views · 100+ GitHub stars","video":"https://www.instagram.com/p/DTSiN8CEiN-/"},{"name":"Scott AI","focus":"AI Coding Assistant","video":"https://www.instagram.com/p/DZzstQwiFNf/"},{"name":"Unify","focus":"LLM Routing Engine"},{"name":"ByteSpace","focus":"Developer Data Platform","video":"https://www.instagram.com/p/DWTrpDQgu6Y/"},{"name":"LangBase","focus":"AI Agent Infrastructure"},{"name":"DigitalOcean","focus":"Cloud Provider","caseStudy":"https://varnan.tech/case-studies/how-we-built-developer-reach-for-composio-digitalocean-and-undermind"},{"name":"Composio","focus":"AI Agent Integrations","caseStudy":"https://varnan.tech/case-studies/how-we-built-developer-reach-for-composio-digitalocean-and-undermind"},{"name":"Undermind","focus":"AI Academic Search","caseStudy":"https://varnan.tech/case-studies/how-we-built-developer-reach-for-composio-digitalocean-and-undermind"},{"name":"Picsart","focus":"AI Media Collaboration","video":"https://www.instagram.com/p/DXzRaCtC7ct/"},{"name":"INC 42","focus":"Startup Intelligence","video":"https://www.instagram.com/p/DZjoIEYGQi9/"},{"name":"SSInnovations","focus":"Medical Robotic Systems","video":"https://www.instagram.com/p/DW-50zEGcw6/"},{"name":"Blue Machine AI","focus":"Deeptech AI Agency","video":"https://www.instagram.com/p/DTpEomkksm0/"},{"name":"Wispr Flow","focus":"Voice AI Writing","video":"https://www.instagram.com/p/DWIrGS6E9ic/"}]},"speaking":{"url":"https://www.parasmadan.in/talks","talks":[{"event":"Inc42 AI Summit","year":"2026","title":"How Consumer AI Products Are Winning Distribution And Retention In India","topic":"Panel · Consumer AI, distribution & retention","source":"https://events.inc42.com/ai-summit/","recording":"https://www.linkedin.com/posts/inc42_inc42-reels-video-activity-7476639819083870208-VYr9","recordingLabel":"Watch AEO clip"},{"event":"TEDxDTU","title":"Becoming a Practical Superhuman: Redefining Creativity with AI","topic":"AI & creativity","recording":"https://www.youtube.com/watch?v=cM_Ebg1yR90"},{"event":"TEDxGGDSDCollege","title":"AI is taking over the world and you don't even know it","topic":"AI & digital identity","recording":"https://www.youtube.com/watch?v=7XuyawcrROs"}],"workshops":[{"organization":"Thapar Institute of Engineering & Technology","title":"AI Agents and the new world we are entering into: Building them from scratch","description":"Led a practical session on building AI agents from scratch.","detail":"Thapar Summer School","date":"14 July 2025","source":"https://www.thaparsummerschool.com/tss-2025"},{"organization":"Shaheed Bhagat Singh College","title":"Generative AI for business, finance, and analytics","description":"Led a Generative AI seminar on applications across business operations, finance, and analytics.","detail":"University of Delhi · Generative AI seminar","source":"https://oldwebsite.sbsc.in/department-societies/"}],"additionalWorkshops":[{"organization":"Mastercard","description":"AI workshops for enterprise and finance audiences, including Dubai."},{"organization":"HSBC","description":"Generative AI sessions for enterprise teams in India and Dubai."},{"organization":"TiE Delhi-NCR","description":"AI Bootcamp on practical tools and using AI for business."},{"organization":"Master’s Union","description":"Applied AI sessions and workshops."},{"organization":"Outskill","description":"Mentoring and teaching across AI engineering, agents, MCP, and AI workflows."},{"organization":"Newton School","description":"AI and technology sessions for developers and learners."},{"organization":"Scaler","description":"AI sessions for the developer and technology community."},{"organization":"Tetr","description":"AI sessions and workshops."}],"workshopHighlights":"https://www.instagram.com/s/aGlnaGxpZ2h0OjE3OTIwMTgzODM4NjMyNDg0?story_media_id=3066356907230730198&stkn=MWoxNmxkNWtxdzlpbw==","hackathon":{"title":"Generative AI Hackathon I Organized","organizer":"Paras Madan","status":"past event","duration":"One week","format":"Remote","theme":"For India, from people of India","description":"A generative AI hackathon organized by Paras Madan with Nas.io. Teams researched real problems, built AI products, and deployed working solutions.","sourceUrl":"https://medium.com/@parasmadan.in/top-13-ai-projects-built-in-indias-first-generative-ai-hackathon-dd2a22956659","sourceContext":"Organizer-authored event recap. Jury affiliations reflect the time of the event.","judges":[{"name":"Gaurav Puri","role":"Security Engineer at Meta","image":"/hackathon/judge-gaurav.png","href":"https://www.linkedin.com/in/gauravpuri19/"},{"name":"Ramsri Goutham","role":"Founder at Questgen.ai & Supermeme.ai","image":"/hackathon/judge-ramsri.png","href":"https://www.linkedin.com/in/ramsrig/"},{"name":"Purvanshii Mehta","role":"Founder Lica, Prev. Microsoft Research","image":"/hackathon/judge-purvanshii.png","href":"https://www.linkedin.com/in/purvanshi-mehta/"},{"name":"Balaji Dhamodharan","role":"AI Researcher & Scholar","image":"/hackathon/judge-balaji.png","href":"https://www.linkedin.com/in/balaji-dhamodharan-91b45615/"},{"name":"Kushagra Pandya","role":"Co-Founder & CTO Vidyo.ai","image":"/hackathon/judge-kushagra.png","href":"https://www.linkedin.com/in/tsaingra/"},{"name":"Vedant Maheshwari","role":"Co-Founder & CEO Vidyo.ai","image":"/hackathon/judge-vedant.png","href":"https://www.linkedin.com/in/vedant-maheshwari/"}],"projects":[{"name":"Clothing Swap","url":"https://drive.google.com/file/d/1Hbo4aFivd1YbmVuCfRzgWo6Mfgs1_RD2/view"},{"name":"Revolutionizing Hiring in India","url":"https://drive.google.com/file/d/1UrfzxW3avenguJ55IuCVWw-Yj0qy-FAx/view"},{"name":"Toxicity Identification and Mitigation in Hinglish Audio","url":"https://www.loom.com/share/fbe863b1c6934b0ba856c14b7036fa68?sid=c842ddec-cd60-482e-a75f-2c623c1f8517"},{"name":"Kids Story Generator Using Their Photos","url":"https://storydev.netlify.app/"},{"name":"Nirmaan AI","url":"https://drive.google.com/file/d/1y_TUsyPOGdCXQ3Xbz_RA_JnjExQe6nDe/view"},{"name":"Meme Generator and Trend Helper","url":"https://drive.google.com/file/d/10fL4jXpf4PtQ3NKrzxhm2Z3RXZ_6EazW/view"},{"name":"Productivity AI Journal","url":"https://drive.google.com/file/d/1MlnGVNvC8Oo6cOAnkZcRG6xOTCTjoBPa/view"},{"name":"Revolutionizing Rural Healthcare","url":"https://drive.google.com/file/d/1EiI-TxOt_i2R_4NyAY-iN_eb9ILZWvPJ/view"},{"name":"Gen AI Athlete","url":"https://drive.google.com/file/d/188eLuqqiTinDUTxy8kFnV0ST5FLafgDA/view"},{"name":"योजना साथी","url":"https://www.canva.com/design/DAGIIC_BNZ4/oMqQ_os0oR_3FKJBIM-jxw/watch?utm_content=DAGIIC_BNZ4&utm_campaign=designshare&utm_medium=link&utm_source=editor"},{"name":"Finance Predictive Modelling","url":"https://drive.google.com/file/d/12N69vTzM9Jcq_ShVC8iYAZMnOqhZZuVW/view"},{"name":"Voice My Resume"},{"name":"Mythological Novel Generator"}],"url":"https://www.parasmadan.in/talks#hackathon"}},"articles":[{"slug":"betterwispr-local-wispr-flow-alternative-mac","title":"BetterWispr: A Free, Local Wispr Flow Alternative for Mac","description":"BetterWispr offers open-source Mac dictation and meeting notes. A closer look at local models, privacy settings, language limits and its early releases.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/betterwispr-local-wispr-flow-alternative-mac/share-image?v=3dd3f906066a","url":"https://www.parasmadan.in/blog/betterwispr-local-wispr-flow-alternative-mac","sourceUrl":"https://www.parasmadan.in/blog/betterwispr-local-wispr-flow-alternative-mac/markdown","sourceFormat":"Markdown/MDX"},{"slug":"claude-managed-agents-dynamic-workflows-limits-pricing","title":"Claude Managed Agents Dynamic Workflows: Limits, Pricing and Setup","description":"Claude Managed Agents now supports dynamic workflows in public beta. How the 1,000-agent limit works, what runs cost, and what developers need to verify.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/claude-managed-agents-dynamic-workflows-limits-pricing/share-image?v=256713c2c6f7","url":"https://www.parasmadan.in/blog/claude-managed-agents-dynamic-workflows-limits-pricing","sourceUrl":"https://www.parasmadan.in/blog/claude-managed-agents-dynamic-workflows-limits-pricing/markdown","sourceFormat":"Markdown/MDX"},{"slug":"elevenlabs-brand-kit-elevencreative","title":"ElevenLabs Brand Kit: How It Works in ElevenCreative","description":"How ElevenLabs Brand Kit stores visual guidelines for AI images and video, where it works, and what teams should check before using it for a launch.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/elevenlabs-brand-kit-elevencreative/share-image?v=81173f6df1af","url":"https://www.parasmadan.in/blog/elevenlabs-brand-kit-elevencreative","sourceUrl":"https://www.parasmadan.in/blog/elevenlabs-brand-kit-elevencreative/markdown","sourceFormat":"Markdown/MDX"},{"slug":"missingno-nous-portal-pricing-api-hermes","title":"MISSINGNO on Nous Portal: Pricing, API and Hermes Agent Setup","description":"A free stealth coding model is available through Nous Portal, with useful API details and an efficiency claim that still needs testing.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/missingno-nous-portal-pricing-api-hermes/share-image?v=65c478219b68","url":"https://www.parasmadan.in/blog/missingno-nous-portal-pricing-api-hermes","sourceUrl":"https://www.parasmadan.in/blog/missingno-nous-portal-pricing-api-hermes/markdown","sourceFormat":"Markdown/MDX"},{"slug":"opencode-1-19-upgrade-opencode-2","title":"OpenCode 1.19 Adds an Explicit Upgrade Path to OpenCode 2","description":"OpenCode 1.19 changes how updates install and adds a direct route to V2, making plugin compatibility and existing workflows worth checking before migrating.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/opencode-1-19-upgrade-opencode-2/share-image?v=b8acc65acbf8","url":"https://www.parasmadan.in/blog/opencode-1-19-upgrade-opencode-2","sourceUrl":"https://www.parasmadan.in/blog/opencode-1-19-upgrade-opencode-2/markdown","sourceFormat":"Markdown/MDX"},{"slug":"openjet-qwen3-8-megakernel-rtx-3090-benchmarks","title":"OpenJet's Qwen3.8 Megakernel: What the RTX 3090 Benchmarks Show","description":"OpenJet packages a faster local runtime for Qwen3.8-27B, with RTX 3090 benchmarks showing 78 tokens per second in reasoning mode and narrow compatibility limits.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/openjet-qwen3-8-megakernel-rtx-3090-benchmarks/share-image?v=320edd4655dd","url":"https://www.parasmadan.in/blog/openjet-qwen3-8-megakernel-rtx-3090-benchmarks","sourceUrl":"https://www.parasmadan.in/blog/openjet-qwen3-8-megakernel-rtx-3090-benchmarks/markdown","sourceFormat":"Markdown/MDX"},{"slug":"qwen-image-2-1-turbo-api-license-8-step-generation","title":"Qwen Image 2.1 Turbo: 8-Step Generation, API Access and License","description":"Qwen Image 2.1 Turbo brings 8-step image generation and editing, downloadable weights and a hosted API. The research license needs a closer look before commercial use.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/qwen-image-2-1-turbo-api-license-8-step-generation/share-image?v=80eb242855ea","url":"https://www.parasmadan.in/blog/qwen-image-2-1-turbo-api-license-8-step-generation","sourceUrl":"https://www.parasmadan.in/blog/qwen-image-2-1-turbo-api-license-8-step-generation/markdown","sourceFormat":"Markdown/MDX"},{"slug":"rea-6-4-android-ai-coding-agents","title":"REA 6.4 Adds Android Inspection to AI Coding Agents","description":"REA 6.4 connects coding agents to Android inspection and existing reverse-engineering tools. How it works, what changed, and where its limits matter.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/rea-6-4-android-ai-coding-agents/share-image?v=3ca3d82c7179","url":"https://www.parasmadan.in/blog/rea-6-4-android-ai-coding-agents","sourceUrl":"https://www.parasmadan.in/blog/rea-6-4-android-ai-coding-agents/markdown","sourceFormat":"Markdown/MDX"},{"slug":"sakana-mass-multi-agent-self-supervision","title":"Sakana MASS: Teaching a Model From Its Own Agent Teams","description":"Sakana AI's MASS improves a 27B model through multi-agent self-supervision, with promising research results and important limits on transfer and reproduction.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/sakana-mass-multi-agent-self-supervision/share-image?v=ddd00edcada1","url":"https://www.parasmadan.in/blog/sakana-mass-multi-agent-self-supervision","sourceUrl":"https://www.parasmadan.in/blog/sakana-mass-multi-agent-self-supervision/markdown","sourceFormat":"Markdown/MDX"},{"slug":"satya-nadella-ai-emergency-brake","title":"Satya Nadella's AI Emergency Brake Puts Control Outside the Model","description":"Nadella's October 10 proposal treats AI agents as insider risks. What external controls, audit trails and a working stop mechanism would require.","date":"2026-10-12","category":"Latest in AI","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/satya-nadella-ai-emergency-brake/share-image?v=61db72796c06","url":"https://www.parasmadan.in/blog/satya-nadella-ai-emergency-brake","sourceUrl":"https://www.parasmadan.in/blog/satya-nadella-ai-emergency-brake/markdown","sourceFormat":"Markdown/MDX"},{"slug":"nobody-knows-which-model-to-pick","title":"Nobody knows which AI model to pick.","description":"GPT-4.1 came after GPT-4.5 and this is where the entire problem starts","date":"2026-10-02","category":"AI & engineering","featured":true,"sample":false,"readingTime":5,"previewImage":"https://www.parasmadan.in/blog/nobody-knows-which-model-to-pick/share-image?v=c1320af5ebf9","url":"https://www.parasmadan.in/blog/nobody-knows-which-model-to-pick","sourceUrl":"https://www.parasmadan.in/blog/nobody-knows-which-model-to-pick/markdown","sourceFormat":"Markdown/MDX"},{"slug":"how-muse-instinct-and-openclaw-work","title":"How Muse, Instinct, and OpenClaw Work","description":"the stack behind next generation of agents and how they work","date":"2026-09-25","category":"AI & engineering","featured":true,"sample":false,"readingTime":5,"previewImage":"https://www.parasmadan.in/blog/how-muse-instinct-and-openclaw-work/share-image?v=60b490d46586","url":"https://www.parasmadan.in/blog/how-muse-instinct-and-openclaw-work","sourceUrl":"https://www.parasmadan.in/blog/how-muse-instinct-and-openclaw-work/markdown","sourceFormat":"Markdown/MDX"},{"slug":"ai-gtm-cant-wait-six-months-to-learn","title":"AI GTM Is Just Fast Experimentation","description":"Inside the new GTM playbook where pricing, distribution and even the customer can keep changing.","date":"2026-09-17","category":"GTM & distribution","featured":true,"sample":false,"readingTime":7,"previewImage":"https://www.parasmadan.in/blog/ai-gtm-cant-wait-six-months-to-learn/share-image?v=65a73392f2a3","url":"https://www.parasmadan.in/blog/ai-gtm-cant-wait-six-months-to-learn","sourceUrl":"https://www.parasmadan.in/blog/ai-gtm-cant-wait-six-months-to-learn/markdown","sourceFormat":"Markdown/MDX"},{"slug":"ai-is-saving-time-why-arent-costs-falling","title":"AI Is Saving Time. Why Aren’t Costs Falling?","description":"Employees feel faster. The company’s costs barely move. What happens to all the time AI saves?","date":"2026-09-16","category":"Enterprise AI","featured":true,"sample":false,"readingTime":7,"previewImage":"https://www.parasmadan.in/blog/ai-is-saving-time-why-arent-costs-falling/share-image?v=745ffd2acdb4","url":"https://www.parasmadan.in/blog/ai-is-saving-time-why-arent-costs-falling","sourceUrl":"https://www.parasmadan.in/blog/ai-is-saving-time-why-arent-costs-falling/markdown","sourceFormat":"Markdown/MDX"},{"slug":"the-rise-of-engineer-led-marketing","title":"The Rise of Engineer-Led Marketing","description":"The strange new jobs appearing between engineering and marketing","date":"2026-09-16","category":"GTM & distribution","featured":true,"sample":false,"readingTime":6,"previewImage":"https://www.parasmadan.in/blog/the-rise-of-engineer-led-marketing/share-image?v=a9e245a2adae","url":"https://www.parasmadan.in/blog/the-rise-of-engineer-led-marketing","sourceUrl":"https://www.parasmadan.in/blog/the-rise-of-engineer-led-marketing/markdown","sourceFormat":"Markdown/MDX"},{"slug":"ai-safety-is-becoming-an-ops-problem","title":"AI safety is becoming an ops problem","description":"The model tried to stop. The system kept it running. What happens when an agent’s safeguards fail?","date":"2026-09-13","category":"AI & engineering","featured":true,"sample":false,"readingTime":9,"previewImage":"https://www.parasmadan.in/blog/ai-safety-is-becoming-an-ops-problem/share-image?v=3215334a7092","url":"https://www.parasmadan.in/blog/ai-safety-is-becoming-an-ops-problem","sourceUrl":"https://www.parasmadan.in/blog/ai-safety-is-becoming-an-ops-problem/markdown","sourceFormat":"Markdown/MDX"},{"slug":"anthropic-threat-report-cheaper-models","title":"Every attack in Anthropic's threat report ran on the cheaper models","description":"The attackers didn’t have access to the best models. So what made them so effective?","date":"2026-09-12","category":"AI & engineering","featured":true,"sample":false,"readingTime":7,"previewImage":"https://www.parasmadan.in/blog/anthropic-threat-report-cheaper-models/share-image?v=91cded1addb9","url":"https://www.parasmadan.in/blog/anthropic-threat-report-cheaper-models","sourceUrl":"https://www.parasmadan.in/blog/anthropic-threat-report-cheaper-models/markdown","sourceFormat":"Markdown/MDX"},{"slug":"building-a-meme-generator-mcp-server","title":"Building a meme generator MCP server from scratch","description":"What changes when Claude can actually use an API? Build a meme generator to see how an MCP server connects conversation to action.","date":"2025-05-25","category":"AI & engineering","featured":true,"sample":false,"readingTime":7,"previewImage":"https://www.parasmadan.in/blog/building-a-meme-generator-mcp-server/share-image?v=222093337055","url":"https://www.parasmadan.in/blog/building-a-meme-generator-mcp-server","sourceUrl":"https://www.parasmadan.in/blog/building-a-meme-generator-mcp-server/markdown","sourceFormat":"Markdown/MDX"},{"slug":"evaluating-naive-and-hybrid-rag","title":"Evaluating naive and hybrid RAG using Weaviate and Athina","description":"Does adding keyword search actually improve your RAG pipeline? Compare retrieved context and generated answers against the same dataset.","date":"2024-11-07","category":"AI & engineering","featured":true,"sample":false,"readingTime":9,"previewImage":"https://www.parasmadan.in/blog/evaluating-naive-and-hybrid-rag/share-image?v=8a5807f5cf94","url":"https://www.parasmadan.in/blog/evaluating-naive-and-hybrid-rag","sourceUrl":"https://www.parasmadan.in/blog/evaluating-naive-and-hybrid-rag/markdown","sourceFormat":"Markdown/MDX"},{"slug":"text-to-sql-without-sharing-database-rows","title":"Text to SQL without sharing database rows","description":"A model can help query your database without receiving its rows. Here is the schema-only approach, and the tradeoffs that come with it.","date":"2024-07-18","category":"AI & engineering","featured":true,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/text-to-sql-without-sharing-database-rows/share-image?v=474a42302048","url":"https://www.parasmadan.in/blog/text-to-sql-without-sharing-database-rows","sourceUrl":"https://www.parasmadan.in/blog/text-to-sql-without-sharing-database-rows/markdown","sourceFormat":"Markdown/MDX"},{"slug":"13-projects-from-our-generative-ai-hackathon","title":"13 projects from the generative AI hackathon I organized","description":"A week, new teammates, and a working AI product. Thirteen teams took on problems across language, healthcare, hiring, and everyday life.","date":"2024-06-20","category":"AI & engineering","featured":true,"sample":false,"readingTime":8,"previewImage":"https://www.parasmadan.in/blog/13-projects-from-our-generative-ai-hackathon/share-image?v=53dee52ebd00","url":"https://www.parasmadan.in/blog/13-projects-from-our-generative-ai-hackathon","sourceUrl":"https://www.parasmadan.in/blog/13-projects-from-our-generative-ai-hackathon/markdown","sourceFormat":"Markdown/MDX"},{"slug":"best-way-to-deploy-open-source-llm-models","title":"Best way to deploy open-source LLM models","description":"Run the model yourself or use a hosted API? The answer depends on what you need to control, and what you can afford to maintain.","date":"2024-05-20","updated":"2026-09-14","category":"AI & engineering","featured":false,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/best-way-to-deploy-open-source-llm-models/share-image?v=b38126b9e23a","url":"https://www.parasmadan.in/blog/best-way-to-deploy-open-source-llm-models","sourceUrl":"https://www.parasmadan.in/blog/best-way-to-deploy-open-source-llm-models/markdown","sourceFormat":"Markdown/MDX"},{"slug":"model-quantization-techniques-use-llms-on-your-cpu","title":"Model quantization techniques: use LLMs on your CPU","description":"What if running a language model did not require an expensive GPU? Quantization changes the memory requirements, but the details matter.","date":"2024-02-28","updated":"2026-09-14","category":"AI & engineering","featured":false,"sample":false,"readingTime":4,"previewImage":"https://www.parasmadan.in/blog/model-quantization-techniques-use-llms-on-your-cpu/share-image?v=03ea350951af","url":"https://www.parasmadan.in/blog/model-quantization-techniques-use-llms-on-your-cpu","sourceUrl":"https://www.parasmadan.in/blog/model-quantization-techniques-use-llms-on-your-cpu/markdown","sourceFormat":"Markdown/MDX"}],"articleSections":[{"id":"all","label":"All","category":null,"url":"https://www.parasmadan.in/blog/all"},{"id":"engineering","label":"AI & engineering","category":"AI & engineering","url":"https://www.parasmadan.in/blog/engineering"},{"id":"GTM","label":"GTM & distribution","category":"GTM & distribution","url":"https://www.parasmadan.in/blog/GTM"},{"id":"enterprise-ai","label":"Enterprise AI","category":"Enterprise AI","url":"https://www.parasmadan.in/blog/enterprise-ai"},{"id":"latest-ai","label":"Latest in AI","category":"Latest in AI","url":"https://www.parasmadan.in/blog/latest-ai"}]},"contact":{"url":"https://www.parasmadan.in/contact","emails":["paras@varnan.tech"],"booking":{"url":"https://cal.com/paras-madan/30min","durationMinutes":30}},"sources":{"homepage":"https://www.parasmadan.in/","contentDirectory":"https://www.parasmadan.in/agents/site.json","openApi":"https://www.parasmadan.in/openapi.json","developerDocs":"https://www.parasmadan.in/developers","portfolioApi":"https://www.parasmadan.in/api/portfolio","sitemap":"https://www.parasmadan.in/sitemap.xml","rss":"https://www.parasmadan.in/rss.xml","jsonFeed":"https://www.parasmadan.in/feed.json"}}