Myntra
Indian Institute of Technology, Delhi
Bengaluru, Karnataka, India
12K followers
500+ connections
About
Activity
12K followers
-
Vikram Garg reposted thisVikram Garg reposted thisIf you wonder why honesty is so rare in govt work, especially in road infra ; this video will give you some idea. Ecosystem is designed in a way to evict the honest uprisings.
-
Vikram Garg shared thisBengaluru has produced India's first AI-powered billboard to display outstanding traffic challans of passing vehicles. The stated goal is to use public visibility as a deterrent. However, the public's response to this initiative has been a powerful lesson in civic priorities. The comment sections of posts celebrating this move are not filled with praise, but with a unanimous, resounding call to fix the city's roads. This brings me to a thought-provoking comment by another fellow Linkedin user, who suggested an alternative billboard: One that displays the number of potholes in the vicinity, the timeline of their existence, and the names of the officials and contractors responsible. The money the have taken for the road. Are corporations better serving society by becoming partners in state-led enforcement, or by championing the transparency and accountability that citizens crave? This isn't just about potholes. It's about the social contract. We see Karnataka's finances move from a revenue surplus a decade ago to a significant revenue deficit today. We, as citizens, contribute nearly 33% in income tax, plus GST on almost every transaction, effectively working for the government for half the year. In return, the focus seems to be on punitive measures rather than improving the core quality of life and infrastructure. Ather Energy's new feature, which maps out potholes on routes for its riders, is a prime example of using technology to solve a real, daily problem for citizens. This is the crossroads for "Tech for Good." Do we use our data and platforms to hold a mirror up to the citizen, or do we use them to hold a mirror up to the system? I believe true innovation lies in empowering the public and demanding better. The companies that understand this will be the ones who truly earn the public's trust. #SmartCities #Bengaluru #CSR #TechEthics #UrbanDevelopment #CitizenFirst #Accountability Ref : https://lnkd.in/gZGXRwYF https://lnkd.in/garvWQ_aVikram Garg shared thisIn school, the fastest way to get attention was to call someone out in class. We tried the same thing on the streets of Bengaluru. Along with Crashfree India and the Bengaluru Traffic Police, we set up India’s first AI-powered billboard at Trinity Circle. It scans number plates and, within seconds, flashes pending challans on a giant screen. You can ignore a push notification. Harder to forget this. A simple experiment. If it makes even a small dent in driver behaviour, it’s worth scaling.
-
Vikram Garg reposted thisVikram Garg reposted thisIt took me 2 hours 20 minutes to drive 11km/7mi home last Thursday at 5PM. The good news is that this was on Outer Ring Road in Bangalore, home to ~400 of the world’s Fortune 500 companies and a large portion of Bangalore’s salaried and income tax paying class. The better news is that this road, which is supposed to join this entire city together, comes to a standstill during rains because of the high number of potholes and lack of drainage. The even better news is that Bangalore has among the highest road tax for cars in India, which means I’ve paid more (18% road tax) for the privilege of sitting in this traffic. The best news is that I can see directly where my 30% + 5% surcharge income tax goes: that Bellandur metro line which started 5 years ago… It’ll definitely be done in 15 years so all the money I pay will be worth it then. Can’t wait for my appraisal this year so I can pay more of my hard-earned money to Bruhat Bengaluru Mahanagara Palike (BBMP) - an organization dedicated to public betterment. Exciting times ahead!
-
Vikram Garg reposted thisVikram Garg reposted thisBrand Bengaluru? It has come to this. Danger is closer than we think it is. While we are safely ensconced in social media, our kids are facing real danger on the roads. It is no longer just about the inconvenience or extra time taken to reach somewhere. Let's not wait for something irreversible to take place, if you know what I mean. This happened today morning at Panathur Road. Last month, one of my colleagues got seriously injured on another road, thanks to a nasty fall while navigating a pothole.
-
Vikram Garg shared thisGiants like Nokia, Kodak, and Blockbuster fell not because they lacked talent, but because they stopped experimenting. Tinkering isn't just a hobby; it's a survival strategy in an AI-driven world. Find out what play crucial roles in adapting to the rapidly changing landscape of today's business environment. #Innovation #Leadership #AI #Experimentation #StrategyThe Tinkerer’s Imperative: Why Experimentation Is the Only Antidote to Corporate DecayThe Tinkerer’s Imperative: Why Experimentation Is the Only Antidote to Corporate Decay
-
Vikram Garg reposted thisVikram Garg reposted thisDelivering fashion and beauty starting in 30 minutes isn’t just a logistical feat—it’s a product and tech challenge we’re proud to have cracked with M-Now. Our teams have been working tirelessly to build the tech that sets a high benchmark for what fashion quick commerce experience should be for our millions of customers. We have just expanded the M-Now offering to Delhi-NCR and Mumbai, providing access to 90K+ trend-first SKUs from over 600 unique brands to our users. This marks a major milestone in how we blend innovation and wide selection with the convenience of super quick delivery, and we are still just getting started... Navneet Agarwal Umesh Naidu Anil Thattil Saurabh Mahajan Ranjith Rameshbabu Hanie S. Ravi Kiran Yerrapragada Ashish Gupta Shubham Rathi Siddharth Shrivastava Aishwarya Mishra Maneesh K Dubey Ankit Dua Amit Sood Shriya Kalra TAPAS SAPRA Vikram K. Vivekanandan Barathi Jubin Kapasi Sayooj Surendran #MNow #Myntra #QuickCommerce
-
Vikram Garg reposted thisVikram Garg reposted thisFor aspiring data scientists who are not from a computer science/software engineering background: AI and software engineering have already cannibalized many traditional data scientist roles. Data science has completely changed in the last 10 years. In 2015, the data scientist role was known as the sexiest job in tech. Back then, the core skills required were: 1. Statistics 2. Machine Learning 3. R/Python, SQL 4. Tableau/Power BI for reporting and dashboards These roles were largely about understanding the business problem, pulling data, building models, and communicating insights. You could build models in Jupyter Notebooks and hand over the results to a business stakeholder or an engineering team. That was enough. But in 2025, this is no longer sufficient. The bar has moved dramatically. Today’s successful data scientists (they’re called machine learning software engineers) also need: • Strong software engineering fundamentals – version control, CI/CD, modular code, testing • MLOps experience – MLflow, Airflow, Docker, cloud platforms (AWS, Azure, GCP) • Production mindset – knowing how to take models from development to deployment • AI/LLM fluency – understanding retrieval-augmented generation (RAG), embeddings, vector databases, and agentic workflows • Data engineering overlap – working with large-scale data pipelines, streaming data, Spark, dbt • End-to-end ownership – from idea to deployment and monitoring If you’re from a non-CS background like me, this can feel overwhelming. But it’s not impossible. What helps: • Mastering the fundamentals of coding and systems design • Building real-world projects that showcase production-grade thinking • Learning cloud tools and open-source stacks • Focusing on solving business problems, not just model performance • Collaborating with engineers and learning from them The “notebook data scientist” is slowly disappearing. But if you’re willing to evolve into a full-stack data scientist – blending business acumen, analytical rigor, and engineering execution – the opportunities are still abundant and exciting.
-
Vikram Garg reposted thisVikram Garg reposted this"You need to speak more in meetings." I remember receiving this feedback years ago. I remembered this recently while reading Taha Hussain's post on loud confidence vs. quiet competence. This is a real issue and it has two sides: 👉 The inner voice (Imposter Syndrome) : Competent individuals often hold back, thinking: “My point will not add much value.” / “Why derail the conversation? I will sort it out offline.” 🔦 This mindset shifted for me when my then manager said: “Meetings are meant for sharing. You are added in the meeting because you add value and If things go sideways, that’s what 'take it offline' is for.” 👉 The outer environment (Culture & Facilitation): Sometimes, meetings reward the loudest voice in the room. Not intentionally but subtly, over time. We say “good point” a few times too often to the same voices. 🔦 This unintentionally sidelines the quietly competent. They disengage and find other ways to contribute often invisibly effective. Leaders play a crucial role on both fronts. 👉 Coach the competent to speak, not just deliver. 👉 Equip meeting facilitators to steer discussions with intention and empower them to amplify sharp voices and interrupt the loud-looping ones. Little more thought and time to not make everyone loud but make every voice count. ********************************************************************* Ranjitha R #rrnotes #leadership #culture #womenintech #voice
-
Vikram Garg liked thisVikram Garg liked thisToday marks my last day at Morgan Stanley after an incredible journey of 8 years and 9 months. I’m deeply grateful for the opportunities, learnings, friendships, and experiences that have shaped me both professionally and personally during this time. A heartfelt thanks to Amit Srivastav, Denis Mustafa, and Shekar Ganesh for guidance, trust, and constant support throughout this journey. And a special thanks to the amazing team — it has truly been a privilege working alongside such talented and wonderful people. As I close this chapter, I look forward to the next phase of my professional journey with gratitude and excitement.
-
Vikram Garg liked thisVikram Garg liked thisThe most dangerous thing AI did to engineers wasn’t take their jobs. It made them forget how to think. There’s a new anti-pattern spreading through engineering teams. When in doubt, call an LLM. Need to check if a string is empty? Prompt it. Parse a date? Prompt it. Validate an email? Prompt it. I get it. AI is fast, it’s capable, and it feels productive. But here’s what nobody’s saying out loud: Every time you reach for an API call to solve a logic problem, you’re renting a brain instead of using your own. And that muscle atrophies. The engineers who will succeed the next 10 years aren’t the ones who know how to prompt best. They’re the ones who know when a 3-line function beats a 25-line AI wrapper and when it doesn’t. AI should handle ambiguity, complexity, and scale. Your code should handle the rest. Use the right tool. Keep the sharp one sharp. Where do you draw the line between “AI is genuinely useful here” vs “I’m just being lazy”? #AI #AgenticAI #LLM
-
Vikram Garg liked thisVikram Garg liked thisOver the last 3 months I have been working on something crazy. An AI agent that is able to automate post-training research at Hugging Face. Introducing ml-intern. It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem. It can pull off insane things: We made it train the best LLM for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink from NVIDIA, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%. In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergency, client and multilingual communication. Then upsampled 50x for training. Beat Codex on HealthBench by 60%. For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on hf.co/spaces, watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously. How it works? ml-intern makes full use of the HF ecosystem: - finds papers on arxiv and hf.co/papers, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on hf.co/datasets - browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data - launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like. Releasing it today as a CLI and a web app you can use from your phone/desktop. CLI: https://lnkd.in/dqX-9k7N Web + mobile: https://lnkd.in/dT8A3iTs And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use. Thanks for help Lewis Tunstall , Leandro von Werra, Yoan Di Cosmo, Henri Bonamy, Carlos Miguel Patiño
-
Vikram Garg reacted on thisVikram Garg reacted on thisI am tired of hearing this - “𝘉𝘶𝘵, 𝘰𝘶𝘳 𝘢𝘯𝘤𝘦𝘴𝘵𝘰𝘳𝘴 𝘥𝘪𝘥 𝘯𝘰𝘵 𝘦𝘢𝘵 𝘮𝘰𝘳𝘦 𝘱𝘳𝘰𝘵𝘦𝘪𝘯..” Technically, that is true. I am pretty sure my grandfather never had a protein shake. But, he never needed to. He had to cycle 5km just to get drinking water. He did not sit for 10 hours. He never went to a pub. He definitely didn’t order from Swiggy. Why do we keep comparing two entirely different lifestyles? Diet is contextual. Always has been. My grandfather ate 10-12 rotis and drank half a litre of milk every day. He could get away with eating a 70% carbs diet because of his daily activity. You and I cannot. Before you eat like your ancestors, Try moving like them first.
-
Vikram Garg liked thisVikram Garg liked thisLogic scales the system, and empathy scales the team. For women in engineering, leadership means both. As we continue our #WomenWhoLead series, spotlighting the women at PhonePe who are redefining leadership, we feature Ranjitha R, Head of Engineering (Backend) today. Ranjitha leads our financial suite of products, ensuring every movement of money is accurate and high-integrity across a half-billion user ecosystem. Her team ensures that while the money moves at speed, the underlying financial integrity remains uncompromised across every stage. Talking about whom she looks up to, Ranjitha adds, “I’m constantly absorbing the sharp acumen of the thinkers I work with - my business partners, engineering peers, and leaders. While I look to them for professional strategy and intellectual growth, it’s the values I learned from my mother that keep me grounded. She has provided the foundational 'why' that allows me to navigate the 'how' of my career.” Her vision for young women in engineering is simple. Build your credibility through your work and your empathy. "A great brain for leadership doesn't have a gender - it just has results." When you dive deep into your subject and see your team as people, you don't just solve the puzzle. You lead the way! Follow the #WomenWhoLead series for more stories of women leading with empathy. #WomensWhoLead #LifeAtPhonePe #InclusiveWorkplace #BelongThrive #WomenOfPhonePe #IWD #Diversity #Inclusion #IWD2026
-
Vikram Garg liked thisVikram Garg liked thisWow. I didn’t know we have to pay ₹3,000 to get our own phone back from the police. 2 weeks ago, I lost my mobile in an auto in Andheri. I kept calling my number again and again. Finally, after many tries someone answered and said the phone is at Vile Parle Police Chowki. I rushed there. Honestly I was so relieved. Anyone who has lost their phone knows the feeling. It’s your work, contacts, messages, banking…everything. So I genuinely thanked the officer for keeping it safe. Then he asked me for ₹3,000. I thought maybe I misunderstood. I asked, “For what sir?” He said very casually: “If you want the phone now, 3000. Otherwise bring the bill and the box and come to court tomorrow.” The phone is 6 years old. I obviously don’t have the bill or the box anymore. I showed my Digilocker ID. I showed my physical ID. I told him the phone is mine. Didn’t matter. Standing inside a police chowki at 11 pm, you don’t feel like arguing. You just want your phone back. In the end I paid because I had no option. Mumbai Police has a very strong image online. The campaigns, the posts, the messaging about honesty and service. But I wish, "the people behind the uniform respected that image as much as the people promoting it." Tagging Greater Mumbai Police because I genuinely want to know if this is the official process for citizens to get their lost phones back. #mumbaipolice #speakup #corruption #systemreality #India
-
Vikram Garg liked thisVikram Garg liked thisEveryone's building on Karpathy's autoresearch for ML. I built one for Bayesian Causal Inference with PyMC. 22 experiments ran overnight. 9 crashed. The agent narrowed the credible interval on the treatment effect by 12% and landed within 1.3% of the true causal effect. I was asleep. The twist: in causal inference you can't score against ground truth. The treatment effect is a counterfactual. So the agent optimized a proxy I designed, written entirely in English. I've been calling this pattern Agentic Data Science. The data scientist doesn't build the model. The data scientist designs the system that builds the model: the objective, the constraints, the domain knowledge, the guardrails. Proof of concept. Friendly datasets. But the pattern works. Link to code in the comments
Experience & Education
View Vikram’s full experience
See their title, tenure and more.
or
Publications
-
Robust 3D Garment Digitization from Monocular 2D Images for 3D Virtual Try-On Systems
WACV 2022
In this paper, we develop a robust 3D garment digitization solution that can generalize well on real-world fashion catalog images with cloth texture occlusions and large body pose variations. We assumed fixed topology parametric template mesh models for known types of garments (e.g., T-shirts, Trousers) and perform mapping of high-quality texture from an input catalog image to UV map panels corresponding to the parametric mesh model of the garment. We achieve this by first predicting a sparse…
In this paper, we develop a robust 3D garment digitization solution that can generalize well on real-world fashion catalog images with cloth texture occlusions and large body pose variations. We assumed fixed topology parametric template mesh models for known types of garments (e.g., T-shirts, Trousers) and perform mapping of high-quality texture from an input catalog image to UV map panels corresponding to the parametric mesh model of the garment. We achieve this by first predicting a sparse set of 2D landmarks on the boundary of the garments. Subsequently, we use these landmarks to perform Thin-Plate-Spline-based texture transfer on UV map panels. Subsequently, we employ a deep texture inpainting network to fill the large holes (due to view variations & self-occlusions) in TPS output to generate consistent UV maps. Furthermore, to train the supervised deep networks for landmark prediction & texture inpainting tasks, we generated a large set of synthetic data with varying texture and lighting imaged from various views with the human present in a wide variety of poses. Additionally, we manually annotated a small set of fashion catalog images crawled from online fashion e-commerce platforms to finetune. We conduct thorough empirical evaluations and show impressive qualitative results of our proposed 3D garment texture solution on fashion catalog images. Such 3D garment digitization helps us solve the challenging task of enabling 3D Virtual Try-on.
Other authors -
-
Teaching DNNs to design fast fashion
KDD fashion 2019
See publication"Fast Fashion" spearheads the biggest disruption in fashion that enabled to engineer resilient supply chains to quickly respond to changing fashion trends. The conventional design process in commercial manufacturing is often fed through "trends" or prevailing modes of dressing around the world that indicate sudden interest in a new form of expression, cyclic patterns, and popular modes of expression for a given time frame. In this work, we propose a fully automated system to explore, detect…
"Fast Fashion" spearheads the biggest disruption in fashion that enabled to engineer resilient supply chains to quickly respond to changing fashion trends. The conventional design process in commercial manufacturing is often fed through "trends" or prevailing modes of dressing around the world that indicate sudden interest in a new form of expression, cyclic patterns, and popular modes of expression for a given time frame. In this work, we propose a fully automated system to explore, detect, and finally synthesize trends in fashion into design elements by designing representative prototypes of apparel given time series signals generated from social media feeds. Our system is envisioned to be the first step in design of Fast Fashion where the production cycle for clothes from design inception to manufacturing is meant to be rapid and responsive to current "trends". It also works to reduce wastage in fashion production by taking in customer feedback on sell-ability at the time of design generation. We also provide an interface wherein the designers can play with multiple trending styles in fashion and visualize designs as interpolations of elements of these styles. We aim to aid the creative process through generating interesting and inspiring combinations for a designer to mull by running them through her key customers.
-
Utility in Fashion with implicit feedback
KDD 2018 - Machine learning meets fashion workshop
-
System for Deduplication of Machine Generated Designs from Fashion Catalog
Trends and Advances in Information Systems and Technologies. WorldCIST'18 2018. Advances in Intelligent Systems and Computing
We propose an unsupervised domain adaptation method to address the problem of image dedup on an e-commerce platform. We present a deep learning architecture to embed data from two different domains without label information to a common feature space using auto-encoders. Simultaneously an adversarial loss is incorporated to ensure that the learned encoded feature space of these two domains are indistinguishable
Other authors -
-
Sales Potential: Modelling Sellability of Visual Aesthetics of a Fashion Product
KDD 2017 - Machine learning meets fashion workshop
Look and feel of a fashion product is difficult to quantify as it is essentially subjective and driven by a host of subtle factors. We formulate a mechanism that grades the look of a product, which we call Sales Potential (SP), that captures visual aesthetics. Our approach normalizes the effects of merchandising factors like discounts, price, list views and brand effects introduced in the e-commerce platform that influence buyers' behaviour.
Other authors -
-
Mastering Social Media Mining with R
Packtpub
See publicationExtract valuable data from your social media sites and make better business decisions using R
-
A Hybrid Framework for Event Detection Using Multi-modal Features
International Conference on Computer Vision
The paper presents novel approach for event detection in sports videos by topic based graphical model learning. The characteristics features defining various sport events are extracted by contextual grouping of low-level video and audio features using topic modeling. Event detection is performed by learning the structure of context based distribution of characteristic features by CRF based graphical model.
Other authors -
-
Mobile Subscriber Fingerprinting: A Big Data Approach
IEEE CloudCom
Mobile advertising campaigns must use subscriber data to target subscribers relevant to the product or service being recommended. However, the matching of recommendations to subscribers is most often inexact, and no straightforward “attribute-value” matching algorithm suffices. Arbitrary matching degrades subscriber experience and results in lower conversion rates.
We described a novel architecture which we call subscriber “fingerprinting”. The proposed system is capable of analyzing…Mobile advertising campaigns must use subscriber data to target subscribers relevant to the product or service being recommended. However, the matching of recommendations to subscribers is most often inexact, and no straightforward “attribute-value” matching algorithm suffices. Arbitrary matching degrades subscriber experience and results in lower conversion rates.
We described a novel architecture which we call subscriber “fingerprinting”. The proposed system is capable of analyzing extremely large volumes of data (in the order of terabytes) using sophisticated large scale distributed Extract-Transform-Load (ETL) operations followed by distributed data Analytics involving statistical models. The insights generated from this process can be used to serve personalized recommendations in real-time. The proposed system uses big data Analytics built over a Hadoop ecosystem, and leverages a private cloud infrastructure for deployment.Other authors -
-
Scene Categorization Using Topic Model based Hierarchical Conditional Random Fields
International Conference on Pattern Recognition and Machine Intelligence
A novel hierarchical framework for scene categorization. The scene representation is defined by latent topics extracted by Latent Dirichlet Allocation. The interaction of these topics across scene categories is learned by probabilistic graphical modelling. We use Conditional Random Fields in a hierarchical setting for discovering the global context of these topics. The learned random fields are further used for categorization of a new scene.
Other authors -
-
Searching OCR’ed text: An LDA based Approach
International Conference on Document Analysis and Recognition
The indexing and retrieval performance over digitized document collection significantly depends on the performance of available OCR. The paper presents a novel document indexing framework which incorporates the digitization process errors in the indexing process to improve the overall retrieval accuracy. The proposed indexing framework is based on topic modeling using Latent Dirichlet Allocation (LDA). The OCR’s confidence in correctly recognizing a symbol is propagated in topic learning…
The indexing and retrieval performance over digitized document collection significantly depends on the performance of available OCR. The paper presents a novel document indexing framework which incorporates the digitization process errors in the indexing process to improve the overall retrieval accuracy. The proposed indexing framework is based on topic modeling using Latent Dirichlet Allocation (LDA). The OCR’s confidence in correctly recognizing a symbol is propagated in topic learning process such that semantic grouping of word examples carefully distinguishes between commonly confusing
words.Other authors -
Courses
-
Computer Vision
-
-
Data Structures
-
-
Embedded Systems
-
-
Human & Machine Speech Communication
-
-
Image Processing
-
-
Operations Research
-
-
Pattern Recognition
-
-
Probability & Stochastic Processes
-
-
Signal Theory
-
Honors & Awards
-
Founder Spirit Team Award, Myntra
Myntra
Awarded for pioneering work on the FWD platform, specifically for developing AI-driven solutions for GenZ trend identification and design automation. 2025
-
Tech L&D Ambassador, Myntra
Myntra
Recognized for outstanding contributions as a custodian and ambassador for the company's learning culture and key Learning & Development (L&D) initiatives
-
Myntra Inaugural Hackathon Winner
Myntra
Secured both 1st and 2nd place in the company's first-ever hackathon, demonstrating exceptional problem-solving and competitive technical skills
-
Summer Undergraduate Research Award
Industrial R & D Unit, Indian Institute of Technology Delhi
Awarded for development of faster Iris Recognition module
Languages
-
English
Full professional proficiency
-
Hindi
Native or bilingual proficiency
Recommendations received
1 person has recommended Vikram
Join now to viewOther similar profiles
Explore more posts
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More