Arnav Garg Email & Phone Number

Palo Alto, California, United States

San Francisco, CA, US

Leading Predibase's ML team.Some of my recent work includes:1. Co-creating Turbo LoRA, a parameter efficient training method to jointly fine-tune for quality and speculative decoding, improving throughput by upto 3.5x for fine-tuned LoRA adapters.2. Developing a mixture of agents synthetic data generation algorithm for supervised instruction tuning that.

San Francisco, CA, US

Building a low-code platform for LLM fine-tuning and inference at scale.At Predibase, I focus on applied machine learning, optimizing fine-tuning workflows, and scaling distributed training and inference for open-source large language models. My work includes implementing reliability mechanisms to ensure training jobs are cost-effective, efficient, and.

San Francisco, CA, US

Sydney, NSW, AU

Building machine learning powered smart features for Confluence and Trello. Some of the things I was responsible for during my time at Atlassian:1. Building models to suggest users and spaces to follow on the Confluence Home Feed2. Content recommendation, both in the form of general recommended content across Confluence as well as related pages at the end.

Sydney, NSW, AU

US

Los Angeles, California, US

Los Angeles, California, US

I co-founded DataRes, UCLA’s first data science and machine learning organization that caters to everyone from undergraduates to PhDs.Website: https://ucladatares.com/Facebook: https://www.facebook.com/ucladatares/Medium: https://medium.com/@ucladatares

Sydney, NSW, AU

I was a part of the Core Machine Learning (CML) team, which is the centralized ML team at Atlassian. I worked on scaling feature generation across product-focused machine learning using self-supervised learning, using ideas inspired by SOTA NLP.

US

I led 4 developers to work on open-source air quality data aggregation services for NASA GMAO and the World Resources Institute (WRI).

US

I was a part of a group of 24 fellows across 5 countries (< 1% acceptance rate).

Austin, Texas, US

As part of Tesla's Low Voltage Controllers (Electronic Systems) team, I identified and resolved a critical flaw in the Autopilot SOC validation manufacturing process, significantly enhancing the robustness testing of Autopilot hardware (HW 2.5 and HW 3.0). I also developed a scalable, low-latency, real-time dashboard to monitor and detect issues in.

Santa Clara, CA, US

As part of Expressive's backend and machine learning teams, I developed and productionized deep learning models (BERT, Transformers, CNNs) for tasks like sentence similarity, metaphor paraphrasing, information retrieval, and context comprehension, improving the accuracy of Expressive's virtual service agent.On the backend, I implemented a feature to bulk.

Fully Remote, OO

As an early employee at Kona (formerly Sike Insights), a Kleiner Perkins-backed startup, I helped build a web application (now a Slack extension) for remote teams to assess EQ compatibility and provide managers with personalized insights to enhance productivity and reduce turnover.I also developed an encryption layer around DynamoDB to ensure secure.

Singapore, SG

I was part of a global team of 25 data scientists across the company.I developed a deep neural network to predict the likelihood of users installing the Starbucks mobile app if shown Starbucks advertisements that significantly improved the Starbucks app installation conversion rate.

Gurugram, Haryana, IN

I built a deep neural network to predict expected flight bookings on the day of departure to help marketing teams better target flights with a large volume of empty seats for promotions and discounts. Through this process, I also discovered several insights to increase profits by analyzing trends in loss-making sectors, competitor routes, partial bookings.

Bangalore, Karnataka, IN

I developed the frontend for Mobisy's digital sales force automation mobile application (Bizom) using AngularJS.

Bachelor Of Science - Bs, Computer Science

Education record