About
Always up for learning new things
Articles by Udayraj
Activity
4K followers
-
Udayraj Deshmukh shared thisReal steel is getting closer to reality! First I thought it's magic of cut scenes and ai videos, but the credible sources and their consistency say it's not fake!Udayraj Deshmukh shared thisEmbodied Avatar: Full-body Teleoperation Platform 🥳 Everyone has fantasized about having an embodied avatar! Full-body teleoperation and full-body data acquisition platform is waiting for you to try it out! #Unitree #HumanoidRobot #Robot #AGI #Teleoperation #EmbodiedAvatar #DataAcquisition #AI #EmbodiedIntelligence
-
Udayraj Deshmukh reposted thisUdayraj Deshmukh reposted thisWe launch Agents on Web today at smallest.ai! With this, you can: 1. Feed your entire website to an AI agent 2. Call/Chat with every customer using AI 3. Customize Agent widget using your brand colors With web agents - you don't wait for customers to sign up onto your website! You proactively speak to them, understand their pain points, and increase your lead-to-sign-up ratio! Sign up now and let us know what you think!
-
Udayraj Deshmukh shared thisI would pay more for simple and multipurpose tools over dedicated machines that need special maintenance. Not sure if this product checks all the boxes for being practically useful and actually cheap over time, but I really like this kind of modular and self explanatory design, but kudos for the idea and the inspiration!Udayraj Deshmukh shared thisThe Robot Learning Company (YC X25) builds affordable AI-powered robot arms to make automation accessible beyond factories. Traditional automation doesn’t work outside of factories. It’s expensive, rigid, and takes months to deploy. And today's next-gen robots usually don’t pay off - they’re slow, costly, and don’t save enough labor to make the math work. TRLC is building low-cost, easy-to-deploy robotic workers that learn new tasks from human demonstrations. By reducing costs and complexity, they make robotic labor accessible to companies that couldn't justify it before— from small manufacturers to service providers and everything in between. Congrats on the launch, Jannik Grothusen! 🚀 https://lnkd.in/gGVsY9aj
-
Udayraj Deshmukh reposted thisUdayraj Deshmukh reposted thisChrome DevTools tip: Gemini can modify and save CSS changes directly to your source! No more jumping between tools! Are you tired of manually copying changes from DevTools back to your editor? Gemini to the rescue! You can now ask Gemini in the AI Assistance panel to: ✨ Update and fix CSS within the Elements panel. 🧪 Instantly test changes live. 💾 Save those changes back to your source files on your computer with a connected DevTools Workspaces folder. Learn more: https://lnkd.in/gcFCYCRg If you connect an existing source folder, this streamlines your workflow like never before. In the Elements panel: - Click "Ask AI" - Ask Gemini to modify your CSS (e.g. center the heading) - Click "Continue" to see the changes live - Expand "Unsaved changes" - Click "Apply to workspace" - Review the diff - Click "Save all" This is big for rapid iteration and efficient CSS development! #ai #programming #softwareengineering
-
Udayraj Deshmukh reposted thisUdayraj Deshmukh reposted this🚀 We're Hiring: Senior Backend Developer (Python) – Immediate Joiner 📍 Location: Noida Join Mobilab, one of India's most innovative AI-driven healthtech startups, where you'll build scalable backend systems, power AI diagnostic engines, and manage secure AWS deployments. ✅ 3–6 years experience ✅ Built systems from scratch ✅ Strong AWS + Docker + API Monitoring Apply now 👉 http://nauk.in/dv75Jp7 #Hiring #BackendDeveloper #PythonJobs #TechJobsIndia #AWS #Docker #FastAPI #Django #HealthTech #StartupJobs #NoidaJobs #AIHealthcare #ImmediateJoiner #Mobilab #SoftwareEngineerSenior Backend Developer (Python) - Noida - Mobilab - 3 to 6 years of experienceSenior Backend Developer (Python) - Noida - Mobilab - 3 to 6 years of experience
-
Udayraj Deshmukh shared thisThis is somewhat nostalgic, as I recall myself presenting my own BTech Thesis project with a robust OMR Checking algorithm in 2019! OMRChecker is still helping people from across the world learn about image processing and build their own projects on top of it. Glad to be maintaining the project 8 years on 😄 p.s. contributions welcome - seeking help to close open PRs and release the v2 branch soon :) #omrchecker #FOSSUdayraj Deshmukh shared thisDear connections, I’m excited to share our graduation project: EvalOptix – Optical Form Scanning & Evaluation System! Together with our advisor Mustafa Yeniad, and my teammates Kerem Batu and Murat Kaynar , we developed EvalOptix, an open-source web-based system that automates the processing and evaluation of optical exam forms. EvalOptix combines OMR (Optical Mark Recognition) and OCR (Optical Character Recognition) technologies to read student answers, extract personal data, store them in a PostgreSQL database, and visualize the results through an interactive web interface. We are grateful to our advisor Mustafa Yeniad for his guidance, and to my teammates Kerem and Murat for their great collaboration throughout this journey. Key technologies we used: Python – core backend development FastAPI – RESTful API for image uploads & data access OpenCV – OMR image processing Pytesseract – OCR text extraction PostgreSQL Global Development Group – storing student answers and OCR data Docker, Inc – containerization and service orchestration Pandas , NumPy – data parsing and analysis Matplotlib, Chart.js – result visualization HTML, TailwindCSS – responsive frontend interface We built this system on the open-source project OMRChecker by Udayraj Deshmukh , which we extended and customized extensively to meet real-world needs.
-
Udayraj Deshmukh shared this**Call for contributors** Hello Everyone! OMRChecker is once again participating in Hacktoberfest in 2024. Project link: https://lnkd.in/gB7Zq3DG What can contributors get from this project? - Great learning experience to up your game in computer vision. - 35+ labelled and organized issues to solve from. - Dedicated Discord channel for doubts as well as open discussions/mentorship in this field - Newcomers can start contributions by submitting their first OMR sample - Visually rich outputs - for learning easily what is happening under the hood. - Working on a well structured large code base that follows pep8/black standards, has pre-commit hooks, and pytest setup ready. - Learn about a structured way of presenting your work via images. - Real life application of basic image processing concepts like edge detections, contour detections, template matching, morphology. - If your PR is accepted, this is a good seasoned project to add to your CV. (#1 for omr tag on github) Please join the Discord server to discuss more: https://discord.gg/qFv2Vqf Cheers!GitHub - Udayraj123/OMRChecker: Evaluate OMR sheets fast and accurately using a scanner 🖨 or your phone 🤳.GitHub - Udayraj123/OMRChecker: Evaluate OMR sheets fast and accurately using a scanner 🖨 or your phone 🤳.
-
Udayraj Deshmukh reposted thisUdayraj Deshmukh reposted thisWe have successfully reached the two-year milestone at FOSS United Bangalore! 🥳 As we celebrate two years of promoting FOSS, let’s take a moment to reflect on our journey so far.🐾 Over the past two years, we've conducted 20 meetups with an audience of over 1600+, and received 300+ proposals! ✨ The talks featured 116 speakers, including 31 women and 10 first-time speakers! 🙌 Thanks to all for your enthusiasm for FOSS and your participation!! 🔥 A big round of applause to the audience whose support helped us reach this milestone! It wouldn’t have been possible without you all! 🙏 The 2nd anniversary special May meetup was a spectacular success! 🥳 A heartfelt thanks to Rippling for providing the venue and refreshments! Special thanks to the speakers Udayraj Deshmukh, Balu Babu, and Dhananjay Ramakrishnappa for their informative sessions! 🔥 Gratitude to the panelists, Anwesha Sen, Shree Kumar, Sai Krishna, Jayasri Vaidyaraman, and Sai Rahul Poruri for their insightful discussion on Public Policies! ✨ A big shout-out to the FOSS United Bangalore team!! 🫶 Aakansha Doshi Shrilakshmi Shastry Vishal Arya Ashwini Manoj Varshini J.m Ishan Sharma Sweta Tanwar Sujal Maiti Vaishnavi Nandakumar Shloka Shah Likhith Prithvi Fenil Jain Yash M. Mangesh Mane Jacob Salvi, Arun Thiagarajan and Yusuf . Huge thanks to everyone who helped us in reaching this milestone! 🌟 Let’s strengthen our efforts to promote the FOSS ecosystem! Meet you all at our next meetup soon!! Rippling #Meetup #Bangalore #May
-
Udayraj Deshmukh reposted thisUdayraj Deshmukh reposted thisSpeaker #2 for the 2nd anniversary special May Meetup! ✨ Join Udayraj Deshmukh, the creator of OMRChecker, to explore its evolution and dive into the journey of image processing with OMRChecker! 🚀 📅 25th May 🕕 11 AM 📍 Rippling Reserve your seat: 🪑 https://lnkd.in/gBt2Vp3N FOSS United #Meetup #Bangalore #May
-
Udayraj Deshmukh liked thisUdayraj Deshmukh liked this📊 OpenFX raised $94M in Series A to solve cross border settlement via 'invisible stablecoin settlements'. Over $45B in annualized volume & 98% of transactions settling in < 60 minutes. 📊 OpenFX has been rebuilding how cross border payments settle, and what is interesting is how they're building it - with stablecoins at the core of their offering. For me - this is super interesting, since for the last 2 years, I've been tracking how stablecoins work, and it's strengthening my belief that stablecoins are the future of both cross border money movement and agentic. machine payments This is how OpenFX works (what I've pieced together, could be some differences). Lets assume the money movement needs to happen from US to Mexico 1. Sender pays fiat in USD via OpenFX API for Mexico transfer 2. OpenFX pays to the Mexican entity from the pre-funded local pool in Mexico 3. Later, OpenFX uses stablecoins internally to rebalance, moving the US value it received from the requesting entity into replenishing the MXN pool. This happens within hours I assume. In a past Painted Stork piece, I had talked about how the ideal flow for stablecoins is backend infrastructure. The customer doesn't know or care how the money is being transferred, as long as it is. They don't care about stablecoins, fiat or any of that jargon. And that's what has happened here. OpenFX themselves call the stablecoin transfer "the smallest step." The real moat they've built is everything around it: ✅ Pre-funded liquidity provisioning across 40+ currency pairs ✅ Last mile payout via local rails ✅ Compliance screening baked into the flow ✅ 90% of trades settling under 60 mins, 30% under 10 mins And this replaces current flows (SWIFT, 2-7 day settlement) with a stablecoin backbone that's invisible to end users. But that is also the thing to remember about the $94M fundraise: 👉 When xborder settlement players raise large rounds, a significant chunk isn't "growth capital" in the traditional startup sense, it's working capital. A large part of this is locked into liquidity pools, and local currency reserves across every supported corridor. It bring to mind the BNPL start-ups of a few years ago: a large part of their fund raise went into FD for FLDG guarantees, not for actual infra / growth investment. And opening new corridors requires new pre-funded pools which = more capital So when you see the headline, temper it. Not all of that $94M is going into growth. A big part of it IS the product itself. The liquidity that makes sub 60 minute settlement possible. This is what "stablecoin infrastructure" looks like. Companies doing this don't need their users to know stablecoins are involved. Money movement just happens. Worth watching. For more pieces like this, check out The Painted Stork (link in comments)👇
-
Udayraj Deshmukh liked thisUdayraj Deshmukh liked thisRecently, we switched from RAFT to our homegrown gRPC-based cluster gossip flow. It’s extremely fast, with minimal consensus overhead. It’s completely peer-to-peer, with no leadership required for replication. Timestamp-based decisions enable a high degree of accuracy in budgeting, rate limiting, and anomaly detection. Releasing with our 1.4.0 series of Enterprise builds.
-
Udayraj Deshmukh liked thisThis moment is a big one for us, but more importantly it reinforces what we’re building toward giving special education teams enterprise grade tools that actually reduce the burden of compliance and help drive better student outcomes. We’re doubling down on that momentum and I am hiring across engineering: - Founding Engineer, Integrations https://lnkd.in/gSTFr2hX - Founding Engineer, AI https://lnkd.in/gUDXsBGa If you’re excited about building at the intersection of AI and education, and want to work on real problems that impact classrooms every day, apply now or reach out to me. #hiringUdayraj Deshmukh liked thisJournify Learning won the K-12 track at the ASU+GSV Summit Cup 50! 🏆 ✨ 3,000+ edtech companies nominated globally → 50 selected to pitch → 1 winner in each category (K-12, Higher Ed, Workforce) We had the honor of taking the stage in front of 1,000+ education leaders. The best part of it all - having special education represented on one of edtech’s biggest stages 💙
-
Udayraj Deshmukh liked thisUdayraj Deshmukh liked thisEngineers hate this, but there's a lot of situations where it can be useful to be less rational for a period of time. So many that "punctuated rationality" is a valid default for many environments. I constantly struggle with this. I desperately want to intellectually "muscle through" tough problems when a better strategy is think deeply, boundedly - and then put on the blinders and execute. It's more aligned with human motivation and you're getting information through execution which is richer than any first-principles model you might build in your head. A few examples: * It's not useful to be constantly evaluating whether you're in the right job. If, like most people, you float between "this is suboptimal" and "this is pretty good", the repeated self-evaluation is just a drag on motivation. Checking in every few months, and doing it deeply, is almost always better. If you're going to look around, don't do it half-assed. * If you're creating a startup you need to be rational and take in all the data around you. But almost all ideas aren't obviously true, so there's long stretches where you need to suspend disbelief and just grind on the idea you decided was worth exploring before you even get real information that tells you whether the direction was correct. * Organizations often set goals and those goals seem less valid as the quarter/half stretches on. But the cost of nitpicking and tweaking the goal is the momentum of the team grinds to a halt. In most cases it's better to power through and establish the cadence. Then move forward with a rhythm of being hyper-diligent upfront and then laser focused through a cycle. There's obviously a large set of people for whom "think a little bit more upfront!" is broadly good advice. But for the rest being able to swap between Spock and Kirk can be a superpower.
-
Udayraj Deshmukh liked thisUdayraj Deshmukh liked thisTraining and real-world analysis shouldn’t be limited by logistics or physical access. Our platform brings the site directly to your desk. We’ve moved beyond the era of just "looking" at models, we are now in the era of active, high-impact interaction. From isolating critical components to accessing deep-dive documentation inside a scene, our platform transforms how professionals train, plan, and analyze. Our interactive Digital Twin technology allows you to explore complex scenarios with total precision. Whether you are in emergency response or heavy industry our platform enables: • Total Immersion: Navigate environments in 3D. • Data Integration: Access PDFs, videos, and equipment specs instantly via smart hotspots. • Precision Tools: Measure distances, isolate components, and compare side-by-side scenarios. • Operational Clarity: A single source of truth for your entire team. We aren't just creating 3D models, we are creating decision-support systems. What scenario could you solve if you had it in the palm of your hand? Let’s talk in the comments. #DigitalTwins #RescueTraining #EmergencyResponse #TrainingInnovation #SpatialIntelligence #Holmatro
-
Udayraj Deshmukh liked thisUdayraj Deshmukh liked thisIf you are trying to understand distributed systems, focus less on what and more on why, when, and how. Answering what a system does is usually the easy part. The harder one is why it behaves the way it does. This is actually learnable, and it comes from building intuition around trade-offs and from spending hours working through that one strange bug. And that intuition is what really matters when things break. In those moments, surface-level or theoretical knowledge is not enough. What matters is having that ability to reason from 'first principles' and apply deep practical understanding. For example, imagine a system where requests suddenly start timing out after a new deployment. Logs look normal, dashboards seem fine, and nothing obvious is broken. If you only know what the system does, you are stuck. But if you understand why it behaves the way it does, you start asking better questions - did latency increase due to a hidden network hop, did retries amplify load, did a small configuration change trigger cascading failures? That is where real debugging begins. So, when you are picking up distributed systems, spend more time asking why, when, and how. Hope this helps.
-
Udayraj Deshmukh liked thisUdayraj Deshmukh liked thisThe next step after Karpathy's LLM wiki idea: . . Karpathy's wiki works on knowledge that sits still. A page on how attention works is just as useful today as it was a year ago. The LLM reads sources, pulls out ideas, writes clean articles, and keeps them cross-linked. You never have to rebuild the context from scratch when you want to ask something. But this breaks the moment you ask a question that spans multiple things at once. "Which authors moved from Google to Anthropic between 2022 and 2024, and what did they publish after the move?" A Markdown page can't answer that. The answer lives in the connections between people, companies, papers, and dates. A wiki can describe that pattern only if someone already wrote an article about it. A graph lets you ask it directly, ask ten variations of it, and get an answer every time without rebuilding anything. FalkorDB is an open-source graph database built for exactly this kind of question. The idea underneath it is simple, and it's what makes the whole thing fast enough to be practical. Most graph databases store connections as chains of pointers and follow them one by one through memory. FalkorDB stores the entire graph as a grid of zeros and ones (a sparse matrix) where a 1 means "these two things are connected." Once your graph is a grid, walking through it becomes math. Two hops is one multiplication. Five hops is five multiplications. That sounds like a small change, but it lets the CPU do work in parallel and reuse decades of math research that nobody had applied to graph queries before. In practice, this is the difference between a seven-hop question returning in 350ms and the same question timing out. The wiki and the graph aren't competing. They sit at different layers. The wiki stores what something is. The graph stores how everything connects. Any work where the connections matter as much as the things being connected belongs in a graph. FalkorDB also comes with vector search built in, which matters for GenAI work. You can find a relevant part of the graph, search for similar items inside it, and return the answer, all in one query. Most GraphRAG setups build this by hand across two separate databases. Here you get it in one. You run it through Docker, query it with Cypher, and connect from Python, JavaScript, Rust, Java, Go, or any Redis client. Open source and multi-tenant by default, so one instance can host thousands of separate graphs without spinning up thousands of servers. Link to the repo in the first comment. Karpathy nailed the foundation. The next layer is here. ____ Share this with your network if you found this insightful ♻️ Follow me (Akshay Pachaar) for more insights and tutorials on AI and Machine Learning!
-
Udayraj Deshmukh liked thisUdayraj Deshmukh liked thisCompanies are revealing what they think their software engineers will do in the future by how they’re changing interviews. I see three broad models emerging. 1. Fundamentals-first. These companies still rely on LeetCode-style coding interviews, sometimes with tighter guardrails to reduce cheating and get a cleaner signal on raw problem-solving ability. Many are also leaning harder on system design. Underneath that is a clear belief: the tools may change, but strong fundamentals will continue to transfer. 2. AI-augmented. These interviews allow AI, but focus on skills like reading code, debugging, testing, and verification. They deprioritize memorization and emphasize judgment and intuition. What that suggests is a shift in what gets valued. Rote recall matters less, but there are still specific capabilities companies expect engineers to own. 3. Pure Builder. These feel more like hackathons than traditional assessments. They test whether a candidate can take ambiguity and turn it into something useful, with product sense, design instinct, and user empathy all in play. That only makes sense if you believe the role itself is expanding from implementer to general software builder. For candidates, this creates a harder question: What do you think your role as an engineer will look like in the future? You cannot really avoid making your own bet. And once you make it, a lot follows from there: which skills to build, which companies to target, and how to prepare. Forcing this question is making a lot of engineers uncomfortable.
-
Udayraj Deshmukh liked thisUdayraj Deshmukh liked thisSearch for Zerodha on Google or the app stores, and the first thing you'll see is ads from our competitors. The only way for us to show up first is to bid on our own brand keyword. So we'd essentially be paying to be visible when a customer is actively looking for us by name, and if we don't, competitors will happily take that spot. What's worse is that ads now show up above and below our own keyword. That tells you everything about where platforms are headed. The only winners here are the app stores who collect the ad spend on top of the commissions they already charge on in-app purchases. And one way or another, this marketing spend eventually flows back to customers in the form of higher prices. We live in a world where everybody keeps talking about disintermediation, but these platforms are some of the most powerful and profitable gatekeepers in history. They sit in the middle and monetise both sides. If you believe in the logic of enshittification, this is just the beginning. Platforms extract as much as they can, for as long as they can, until they can't. More ads means more scrolling before you hit any organic results, which means brands that don't pay simply stop being visible. That's manageable if you're a large company with a marketing budget. For smaller businesses and startups it can be devastating. They just can't afford to keep up. The sheer irony of having to pay to show up when someone is already searching for you by name never stops being absurd.
Experience & Education
-
Razorpay
View Udayraj’s full experience
See their title, tenure and more.
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Volunteer Experience
-
Volunteer
Guwahati Half Marathon
Health
-
City Representative, Technothlon 2016
Technothlon, IIT Guwahati
Successfully conducted the exam in my city with a participation of over 200 students.
-
Mentor
Indian Institute of Technology, Guwahati
- 1 year 1 month
Social Services
Student Mentor for Peer Mentorship program under SAATHI Club, IIT Guwahati.
Projects
-
Online Shopping System
See projectDeveloped a multi-client ordering system that allows for simultaneous orders to
be processed using synchronization. Built using: Java threading, networking and Swing -
Speech Recognition Software
- Present
See projectWrote speech recognition program in C++ fr om scratch using various speech modelling techniques
-
Mobile-based collaborative OMR Checker
- Present
See projectAn OMR checker that can handle images tak en at any angle or distance from a mobile camera.
Aims to eliminate need of a scanner by allowing multiple devices to scan and process simultaneously. -
Automation for No Dues Forms
A Web App for automating No Dues Clearance Form in IIT Guwahati. It involved an hierarchy of roles so that the requests are filtered according to level of the role.
Other creators -
-
Smart Gardening System
-
Created a raspberry pi controlled mini-garden with automated irrigation and sunlight control with live
monitoring of temperature, light amount, etc via MQTT protocol. -
Rubik's Cube Solver
-
See projectMade an autonomous robot that solves a physical Rubik's cube. This is a project under Robotics Club, IIT Guwahati.
-
Fluxo-de-Energia: Technothlon online event
-
See projectA visually attractive web-based multiplayer game inspired from virtual stock
market. Enables users to play different roles and experience stock market in a fun way.
Built using: Laravel 4 (PHP), jQuery Ajax, mySQL -
Technothlon Android App
-
See projectContributed to the backend development for the app based on cordova engine and assemble page generator.
Key Technologies: Cordova, jQuery, Ajax, Assemble
Recommendations received
-
LinkedIn User
7 people have recommended Udayraj
Join now to viewExplore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content