Skip to content
View Siddharth-R512's full-sized avatar
10 followers · 7 following

Achievements

Achievement: Pull SharkAchievement: Quickdraw

Achievements

Achievement: Pull SharkAchievement: Quickdraw

Block or report Siddharth-R512

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
Siddharth-R512/README.md

hey, I'm Siddharth 👋

I'm an AI engineer. I build agent systems that drive real API surfaces through typed tools, and retrieval systems that stay grounded in real data.

The part I care most about is measurement. Most LLM work has no ground truth, so it's easy to ship something that feels good and impossible to say whether it's right. I spend a disproportionate amount of time on eval harnesses strict enough to catch a result that looks like an improvement but isn't -- including my own.

: what I work on

  • 🧩 MCP & agent orchestration -- exposing existing APIs as typed tools, scoped tool catalogs, one-agent-per-domain routing to keep selection accuracy stable as coverage grows
  • 🔍 Retrieval -- structural chunking, hybrid BM25 + RRF, version-aware filtering
  • 📏 Evaluation -- frozen golden sets, deterministic LLM-free scoring, pre-registered predictions, per-question analysis instead of aggregate-only reporting
  • 🔒 Structured output -- schema validation in JSON mode, retry on violation, pinned seeds for reproducible runs

I build directly against SDKs rather than reaching for a framework, mostly so I can see what's actually happening in the retrieval and tool-call paths.

: stack

Python · FastAPI · Pydantic · MCP (FastMCP) · Qdrant · OpenAI / Azure OpenAI · SQL · Git

: connect

Happy to talk about retrieval, agent design, or how you're evaluating any of it.

Pinned Loading

  1. Knowledge-Based-QA-using-DistilBERT Knowledge-Based-QA-using-DistilBERT Public

    Implementing Knowledge-Based Question Answering (QA) leveraging DistilBERT, enhancing information retrieval accuracy.

    Jupyter Notebook

  2. Martian-Terrain-Segementation-and-Analysis Martian-Terrain-Segementation-and-Analysis Public

    Applying deep learning techniques to partition Martian terrain, facilitating precise identification and characterization of geological formations for planetary research.

    Jupyter Notebook

  3. Online-Boba-Tea-System Online-Boba-Tea-System Public

    Java

  4. PES2UG21CS524_Jenkins PES2UG21CS524_Jenkins Public

    C++

  5. Small-car-page Small-car-page Public

    Applying concepts i have learnt so far.. till fb.

    CSS