The Fifth Elephant 2020
2020-06-19..2020-06-20
-- online --
[The Fifth Elephant 2013, The Fifth Elephant 2015, The Fifth Elephant 2016, The Fifth Elephant 2018, The Fifth Elephant 2019, The Fifth Elephant 2020, The Fifth Elephant 2023 Monsoon, The Fifth Elephant 2025 Winter Edition, The Fifth Elephant Pune edition 2026]
| title | author | length |
|---|---|---|
| Real-time Machine Learning Inference Platform | Nikunj Jain | 20m |
Data scientists and ML engineers work on a variety of problems at Zomato such as predicting kitchen preparation time, predicting rider assignment time… | ||
| 7 steps to build your own data pipeline | Kumar Pushpesh | 41m |
CTO and co-founder of Moonfrog Labs - Kumar Pushpesh - explains why the company built data infrastructure in parallel to games/products, including:
1. Having… | ||
| Building a scalable data platform | Jayesh Sidhwani | 32m |
In this talk, Jayesh Sidhwani - former head of the data infrastructure at Hotstar - shares the patterns learned in the process of building a scalable data… | ||
| Capacity and efficiency problems in real-time logistics - and how data at Swiggy solves this | Piyush Srivastava | 41m |
In this talk, Piyush Srivastava - Director of Engineering for the Delivery team at Swiggy - discusses the nature of historical and real-time data and Swiggy's… | ||
| Detecting anomalies in Flipkart's fulfillment network | Govind Pandey | 29m |
Flipkart moves from dashboards to an automated anomaly pipeline that samples metrics, stores them in an incremental data platform, and scores anomalies with… | ||
| Introduction to cryptography | Anand Venkatanarayan | 01h 07m |
Traces a concrete crypto timeline showing how Amazon S3 moves from client-side encryption (2006) to server-side, to KMS and bring your own keys (2011–2016)… | ||
| How to build your data engineering/data science team? | Goda Ramkumar, Srujana Merugu | 01h 05m |
Explains how to structure teams: Series B startups run a central data science team while mature companies shift to a hub-and-spoke model with embedded analysts… | ||
| DevOps for Machine Learning Projects | Dmitry Pretrov, Ivan Shcheklein | 59m |
Argues that treating datasets and models as versioned artifacts (not just code) makes ML reproducible and promotable through CI/CD, and shows DVC workflows… | ||
| Machine Learning Models and Productionization | Aditya Patel, Nischal HP | 01h 05m |
Few models reach production because real constraints bite: on‑premise insurance deployments require shipping to private clouds with no access, training data is… | ||
| Experimentation in Data Science | Paul Meinshausen, Bargava Subramanian | 01h 03m |
Argues that experimentation is the systematic way to learn in production, not just A/B tests, and shows concrete practices: use multi-armed bandits for… | ||
| Doing Data Science with Ethics | Suchana Seth | 01h 01m |
Argues for concrete rules: register ethics goals at design time, track data provenance and consent, and treat fairness, interpretability and robustness as… | ||
| Reducing the cost of production AI | Venkata Pingali | 26m |
Experience shows feature extraction across billions of event rows runs in three tiers (weekly batch, daily/near‑time, minutes) and can be a 12‑hour job that… | ||
| Data Science for Financial Inclusion | Thuong Nguyen | 31m |
Trusting Social replaces bank histories with telco call/SMS/top-up logs and combines other alternative sources to score the unbanked, claiming coverage of… | ||
| Data governance: lessons on data usage and data controls from finance domain | Kaushik Das | 07m |
After regulators demand numbers in 2016, a Wells Fargo data program builds a pipeline that detects PII at ingestion using a data scan mechanism, applies… | ||
| On financial modelling, credit scoring, and risk assessment | Thuong Nguyen, Chris Stucchio, Anand V | 01h 19m |
Explains how lenders turn social, telecom and transaction traces into usable scores: extract ~10,000 features, reduce to ~100–200, train models with… | ||
| On Kalman Filters and Robotics | Naman Kumar | 43m |
A working agricultural robot fuses camera-based deep learning weed detections with a Kalman-filtered navigation stack that combines IMU and wheel-encoder… | ||
| Robust NLP Models for real-world deployment | Sandya Mannarswamy | 36m |
In this talk, speaker Sandya Mannarswamy argues that one of the reasons why many NLP models don’t generalize and fail in real world is due to the lack of… | ||
| Inference algorithms and sampling techniques | Vinodh Kumar | 26m |
A lot of unsupervised learning algorithms work by inferencing parameters of generative models through Monte Carlo techniques.
In this talk, speaker Vinodh… | ||
| Deep Learning and Optical Character Recognition | Nischal HP | 40m |
When you think about document digitization from a business optimization process perspective, simply performing OCR does not solve the problem. In this talk… | ||
| AI and gendered biases | Radhika Radhakrishnan | 30m |
Field study tests one device each for five assistants (Siri, Google Now, Bixby, Cortana, Alexa) with a standardized set of English prompts—three sexual… | ||
| ML application lifecycle | Srujana Merugu | 43m |
Building good ML systems is not very unlike developing good software. Just as developing good software requires mastering not only programming theory, tools… | ||
| Data labelling to scale your ML and AI pipelines | Bikram Sengupta | 40m |
In Software 2.0, data is code. A mindful approach to your data annotation pipeline and practices is critical to the outcomes of your ML algorithms. If not done… | ||
| On Tensors | Nishant Sinha | 28m |
Current tensor libraries leak low-level flat-memory semantics so reshape and broadcasting rules quietly produce wrong results; the claim is that naming… | ||
| Feast: feature store for Machine Learning | Willem Pienaar | 27m |
GO-JEK builds Feast with Google Cloud tooling to stop feature drift and speed production ML: a Python SDK, gRPC online serving and BigQuery warehouse, later… | ||
| Why Mozilla is focussing on better data | Rebecca Weiss | 28m |
Claim: a privacy-first culture yields higher-quality, cheaper data. Evidence: 26% of companies report inaccurate data, only 35% have centralized stores, and… | ||
| Lean Data as competitive advantage for business | Stan Leong | 23m |
Argues that lean data is a competitive advantage not overhead, backed by concrete breach examples: $5 billion fine for Facebook/Cambridge Analytica, 145… | ||
| Q&A session -- The Big Data Paradox public lecture | Rebecca Weiss, Urmika Devi Shah, Stan Leong | 39m |
Q&A documents concrete Lean Data practices: prioritize in-context notices over long policies, run timed experiments (start with 1% experiments for six months)… | ||
| Creating responsive data visualisations | Gurman Bhatia | 01h 30m |
Ten concrete methods for making visualizations work across screens, from design for the phone first to choosing static images over costly animation; claims… | ||
| nothing matches | ||