Skip to content
View sauravsingla's full-sized avatar
🎯
Focusing
🎯
Focusing
51 followers · 0 following

Achievements

Achievement: Pull Sharkx3Achievement: StarstruckAchievement: YOLOAchievement: Quickdraw

Achievements

Achievement: Pull Sharkx3Achievement: StarstruckAchievement: YOLOAchievement: Quickdraw

Block or report sauravsingla

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
sauravsingla/README.md

Saurav Singla

Building open-source systems for AI agents, efficient LLM inference & dynamic graphs

Agentic AI · LLM Systems · Graph Analytics · C++ / Python

LinkedIn Google Scholar Hugging Face ORCID

01 — 🧠 AgentWeave

Route before you reason.

Pre-inference routing and secure execution for tool-rich LLM and multi-agent systems. AgentWeave reduces the action space shown to a model before inference while keeping authorization, provenance and recovery explicit.

Project benchmark: 70.18% fewer tools exposed · 61.70% fewer input tokens · 50.95% lower mean local-model latency

MCP · A2A · LangGraph · AutoGen · Python

Repository · Project site · Paper

02 — ⚙️ MemVanta

Run quantized LLMs with less resident memory.

A memory-first C++20 runtime for quantized GGUF models on CPU, built around mmap-backed model access, paged KV cache, compact kernels and reproducible benchmarking.

7B benchmark: 3.80 GiB peak RSS · 47.54% lower peak RSS than the pinned comparison runtime

C++20 · GGUF · CPU Inference · Systems

Repository · Docs · Benchmark

03 — 🕸️ VeloGraphX

High-performance analytics for continuously evolving graphs.

A C++20 + Python engine for dynamic graph analytics with adaptive repair vs recomputation across BFS/SSSP, connected components, triangle counting, k-core and PageRank.

Retained exactness stress result: 2,000,000 updates · 0 BFS mismatches · 0 triangle mismatches

C++20 · Python · Dynamic Graphs · Graph Analytics · PyPI

Repository · Docs · PyPI

Open-source approach

I try to make projects useful beyond a demo: benchmarks, reproducible evidence, tests, releases, documentation and explicit claim boundaries alongside the code.

If you work on agent systems, efficient inference, graph analytics or production ML, I’m happy to compare notes and collaborate.

Research & Publications

Google Scholar · IEEE Xplore · ORCID · DBLP · OpenReview · ACM DL

Book: Machine Learning for Finance · Course: Data Analysis for Business and Finance · Technical writing: Towards Data Science, HackerNoon and KDnuggets

Research highlights · Research impact · Open-source contributions · Industry recognition

Build → Measure → Publish → Improve

Follow @sauravsingla for new releases, benchmarks and open-source experiments.

Pinned Loading

  1. agentweave agentweave Public

    Pre-inference tool and agent routing for MCP, A2A and tool-rich LLM systems. Reduce candidate tools before model inference with policy-aware routing and recovery.

    Python 39 36

  2. MemVanta MemVanta Public

    Low-memory C++20 LLM inference runtime for quantized GGUF models on CPU — mmap-backed weights, paged KV cache, Q4/Q8 kernels, and reproducible llama.cpp benchmarks.

    C++ 39 38

  3. VeloGraphX VeloGraphX Public

    High-performance C++20 + Python engine for dynamic and incremental graph analytics on evolving graphs — BFS, SSSP, connected components, triangles, k-core and PageRank.

    C++ 36 33

  4. DeciShift DeciShift Public

    CPU-first framework to compare ML decision-system versions, identify changed decisions, and attribute shifts to features, models, thresholds, rules, and interactions.

    Python