Skip to content
View ashwinsreedhar28's full-sized avatar
⚡
⚡
3 followers · 3 following

Achievements

Achievement: Pull Sharkx3Achievement: Pair Extraordinairex4Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pull Sharkx3Achievement: Pair Extraordinairex4Achievement: QuickdrawAchievement: YOLO

Block or report ashwinsreedhar28

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ashwinsreedhar28/README.md

Hi, I'm Ashwin 👋

Backend & AI infrastructure engineer, finishing an MS in Computer Engineering at Purdue (Dec 2026).

What I've built

  • emberserve: LLM inference engine from scratch (Python + Triton). 23% → parity with vLLM on a 7B through nine profile-driven fixes; Qwen3-8B cold start 6.7 s vs vLLM's 69 s.
  • serverless-lakehouse: medallion lakehouse over GPU cold-start data from Runpod Serverless. Built on PySpark + Delta, ported to Snowflake + dbt with 8/8 gold tables at parity; live Runpod API poller; dataset + dashboard on Hugging Face.
  • Pulse: local-first macOS news intelligence app with local and Claude LLM routing.

Open source: compile-cache fix merged into Runpod's vLLM worker (v2.29.0). Caches torch.compile output on network volumes, cutting ~40–60 s per cold start.

Before: platform developer at E&J Gallo: production LLM agents and an automation fleet running 30–50 jobs a day.

Open to new-grad backend / AI infra roles starting January 2027 · LinkedIn

Pinned Loading

  1. emberserve emberserve Public

    From-scratch LLM inference engine — continuous batching, paged KV cache, prefix caching, hand-written Triton paged attention, CUDA graphs, OpenAI-compatible API. Benchmarked against vLLM on an A100.

    Python 1

  2. serverless-lakehouse serverless-lakehouse Public

    Medallion lakehouse (bronze → silver → gold) over my own Runpod Serverless LLM benchmark data: cold starts, FlashBoot, worker-vllm boot anatomy, cost per job. Local PySpark + Delta Lake, redacted l…

    Python

  3. Pulse Pulse Public

    Pulse — a private, local-first macOS news & intelligence dashboard. Electron + React + SQLite, Ollama-powered urgency scoring, per-team sports alerts, Stooq quotes, Hyperintelligence feed discovery…

    TypeScript 3

  4. Aether Aether Public

    Personal-OS substrate. Holographic Electron shell on a signed mesh, with voice as a first-class participant. macOS, pre-1.0.

    TypeScript 3 1