Bengaluru, Karnataka, India
At Lex AI , I'm building *Prof* (https://prof.lexailabs.com) first-principles ML for the next 10 million engineers. → Reverse-engineer YouTube's recommender, layer by layer. → Train GPT in under 300 lines, on your machine. → Solve hundreds of LeetCode-style ML problems. → Crack that ML interview at Big Tech. Before Lex AI: ex-Google (Cloud Security & AI products, taught ML to fellow Googlers), ex-MathWorks (autonomous driving, speech). CS Engineering in Machine Learning & AI (IIIT Delhi). I write at purukathuria.com and ship deep-learning talks on YouTube. → Try Prof: https://prof.lexailabs.com/ → Newsletter: https://www.purukathuria.com/
Having fun building *Prof* and other interesting products to solve AI Education & AI Adoption with an engineered pedagogy. Solving for Higher Education Institutions, Enterprises & Working Professionals. Prof: > leetcode for deep-learning > learn DL & modern AI to crack machine learning engineering interviews at big tech & AI Labs.
Automated Driving, Backend Core Responsibilities: Responsible for leading design and implementation of upcoming products around scenario variant generation. Transitioning requirements to MVP to shipping product. Responsible for converting an idea to design spec, brainstorming it with cross functional teams to implementing, testing and productizing it. Other Activities: Led a Deep Learning book club. Interviewing (on campus and off campus for Engineering roles) Mentoring new hires/juniors on multiple fronts. Bashing/Bug Finding Events. Hackathons. -------------------------------------------------------------------------------------------- Signal Processing and Deep Learning, Backend Implemented Keyword spotting Algorithm, machine fault detection algorithm and deployed it on Intel-MKL DNN and Raspberry-Pi. MATLAB to C++ for deployment. Feature Engineering on Speech signals. Bi-LSTMs and CNNs. -------------------------------------------------------------------------------------------- Languages: Java; C++; Python; MATLAB; SQL; Theoretical Computer Science: Algorithm Design; Signal Processing; Machine Learning; Deep Learning; Linear Algebra; Image Processing; Motion Planning for Self driving cars;
Automated Driving, Backend. Algorithms for scenario generation and simulation: Reduced search space and improved the Actors field of view algorithm's runtime by 75%. Represented the road network as a Graph. Scaled the design to Automatic scenario generation. Solved more than 50 real world customer centric problems around ML, DL, Image processing and programming fundamentals. Facilitated and presented knowledge sharing meetings that revolved around cutting edge tech. (DL, Autonomous driving, RL)
APTOS 2019 Blindness Detection on Kaggle under Bio-Informatician and Head of CB at IIIT Delhi, GPS Raghava. (https://en.wikipedia.org/wiki/Gajendra_Pal_Singh_Raghava) Developing Deep Learning Algorithms for detecting diabetic retinopathy to stop blindness before it's too late. Problem: Multi Class Image Classification problem. After performing Data Cleaning, Preprocessing and Data Exploration, experimented with various Algorithms to perform as well in the challenge as possible. (CNNs, DenseNet, EfficientNet, Ensembles of multiple learners) Theoretical Tech: Image Processing; Machine Learning; Deep Learning; Linear Algebra; Calculus. Applied Tech: Python; Keras; PyTorch; Keras; Jupyter Notebook. https://www.kaggle.com/purukathuria/competitions?tab=completed
Areas of Research: Knowledge Graphs, RDF Graphs, Property Graphs, Ontologies, Graph Databases, Triple Stores. - Designed and Developed Algorithms for Evaluating Resource Description Framework (RDF Graphs) and Property Graphs on Triple Stores and Graph Databases. Evaluation was performed using Industry standard benchmarking standards (WatDiv, SP2B and LUBM) from the perspective of Loading, Querying and Reasoning Support. - Evaluated on Neo4j, Cassandra, BlazeGraph, Apache Jena. - Translated all the SQL queries to Cypher Query Language for Neo4j. https://github.com/kracr/rdf-pg-eval/tree/master/Puru.neo4j