Hi, I'm Amogh! I am a recent graduate from
University of California, San Diego, with a M.S. in Computer Science, with an AI/ML specialization.
I'm also currently a growth strategist for
Perplexity AI, and Co-President of the UCSD CSE Graduate Student Council.
I've most recently worked at
AMD, as an Generative AI Applications Development Intern on the RyzenAI team, optimizing and becnhmarking LLM inference on hardware. I've previously worked as a generative AI intern at Marvell, building a gen-AI based synthetic data generation pipeline, and using reinforcement learning and deep learning for DSP parameter optimization.
I've also worked at Amazon Web Services as a software development engineering intern on their AI-powered automatic speech recognition team,
Lex ASR.
I graduated with a B.S. in Data Science from the
University of California, San Diego in 2023. For more
details, check my
CV or shoot me an
email!
I enjoy playing and watching football, basketball and cricket, as well as making latte art, hiking, and (very competitively) playing board games!
AI Applications Development Intern
September 2024 - December 2024
- Optimized generative AI workload (Llama-2, Llama-3, Stable Diffusion, etc.) execution on RyzenAI neural processing unit (NPU) using strategies such as caching, batching, quantization, as well as model and data parallelism.
- Benchmarked generative AI workloads on Ryzen and competitor hardware (Snapdragon X Elite, Lunar Lake), specifically time to first token, and tokens/sec, i.e. latency and throughput, as metrics.
- Evaluated NPU latency and utilization during concurrent execution of generative AI workloads and high-fidelity AI effects (Microsoft Studio Effects).
Data Scientist Intern, Marvell Inc.
Jun 2024 - September 2024
- Architected an end-to-end ETL pipeline consisting of data engineering, analysis and visualization using numpy, pandas, sklearn, Tableau, etc. Data was ingested using Apache and AWS Kinesis, and stored in SnowflakeDB, and AWS S3.
- Utilized generative pretrained transformer (GPT) models to generate synthetic data, leveraging parameter efficient fine-tuning (PEFT) techniques such as low-rank adaptation (LoRA).
- Implemented reinforcement learning (RL) and deep learning methods for DSP parameter optimization.
- Integrated large language models (LLM) and retrieval augmented generation (RAG) to automate hardware modeling process.
Researcher, Stanford University School of Medicine
June 2023 - June 2024
- Performed data engineering, analysis, and visualization through various methods in Python and R; specifically, pandas, geopandas, numpy, matplotlib, scikit-learn and ggplot. Developed unique statistical packages composed of chi-squared and Fisher tests.
- Led research teams mentored by Dr. Eric Gross, Dr. Latha Palaniappan, and Jin Long. Opioid overdose research published in British Journal of Anaesthesia, and diabetes research in preprint at Journal of Asian Health.
Software Development Engineering Intern, Amazon Web Services
June 2022 - Sept 2022
- Led architectural changes in Lex ASR (Automatic Speech Recognition) Services and DataHub, improving latency for conversational AI models. Conducted A/B tests to evaluate architectural changes for customer use.
- Performed cohort analysis to segment critical and non-critical customer data in DataHub, enhacing Lex ASR schemas, and enabled compliant storage of all data using AWS Kinesis, S3, and Lambda.
- Applied time series analysis on AWS CloudWatch metrics to track and optimize the performance of ASR Service, leading to lower response time and accelerating customer request resolution by up to ∼75%.
Research Intern, Scripps Research Translational Institute
Jun 2021 - Aug 2021
- Developed an R library to estimate genetic regulatory variation using a confidence interval estimation method.
- Implemented various statistical concepts like binomial distributions and parametric bootstrapping, and applied them to data
from the Genotype Tissue Expression Project (GTEx).
Firmware Embedded Engineering Intern, Inphi Corporation (acq. Marvell Inc.)
July 2019 - Sept 2019
- Developed Python modules and a user interface to create and display data based on test options and feature selection.
- Developed firmware in C++ to parse data, removing duplicates based on timestamp & hex value, per modular architecture.
Jacobs School of Engineering at University of California, San Diego
M.S., Computer Science and Engineering, AI/ML Concentration
September 2023 - March 2025
Halicioglu Data Science Institute at University of California, San Diego
B.S., Data Science
September 2020 - March 2023