About
Experience
Projects
Blog

Categories

All (4)

benchmarking (2)

BM25 (1)

concurrency (1)

deep-learning (1)

distributed-training (1)

fine-tuning (1)

fsdp (1)

go (1)

inference (1)

llm (3)

modal (1)

moe (1)

Opensearch (1)

pytorch (1)

ray-serve (1)

redis (1)

Semantic Ids (1)

systems-design (1)

vector search (1)

vllm (1)

Blog

Fitting a 3B Model on Two GPUs with FSDP: A Practical Walkthrough

fsdp

distributed-training

pytorch

llm

fine-tuning

modal

These are my notes from fine-tuning SmolLM3-3B on a medical Q&A dataset across two GPUs using FSDP (Fully Sharded Data Parallel). The goal was practical: understand what…

Rediska: Architecture and Performance Notes for a Go-Based Redis Clone

go

redis

concurrency

systems-design

benchmarking

Rediska is a Redis clone I’ve been building in Go, mostly as a way to actually understand what Redis does under the hood — RESP parsing, an in-memory store, AOF persistence…

Building and Evaluating a Search Engine Over Malware-Classification Papers

A capstone project for AI-Powered Search: BM25 to generative retrieval

llm

BM25

Opensearch

Semantic Ids

vector search

This project builds a search engine over a corpus of arXiv malware-classification papers, implementing and comparing seven retrieval methods head-to-head: BM25 → dense…

Building a Production-Grade LLM Inference Stack: Benchmarking vLLM, Ray Serve, and MoE Models

llm

inference

vllm

ray-serve

moe

benchmarking

deep-learning

These are our notes from benchmarking a self-hosted LLM inference stack. The goal was practical: figure out which combination of model, serving framework, and infrastructure…

Vinay Pandya, Yiran Xu

No matching items

© 2025 Vinay Pandya