8000 bajpainaman (Naman Bajpai) · GitHub
[go: up one dir, main page]

Skip to content
View bajpainaman's full-sized avatar
🏠
Hanging by a thread
🏠
Hanging by a thread

Block or report bajpainaman

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. 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
bajpainaman/README.md

Naman Bajpai. Co-founder and CTO of The Singularity. Markets, systems, and machine intelligence.

Website: namanbajpai.com LinkedIn: bajpainaman X: @defibagholder Email: namanbajpai23@gmail.com Resume (PDF) GitHub followers

I am the co-founder and CTO of The Singularity (Entropy to Order), and I study Computer Science at Drexel University.

My work spans exchange infrastructure, GPU computing, distributed systems, compilers, and applied machine learning.

Building The Singularity

We are building the Everything Exchange: a common system for trading across centralized exchanges and on-chain markets.

My work here spans OneBook for order management, routing, and execution; Hilsa for cross-chain observation and verification; and the matching, consensus, custody, and settlement infrastructure underneath them.

Execution, consensus, and market design
  • Execution and shared state: GPU execution, shared-state systems, and tooling across multiple VM families. The ETO CLI documents EVM, SVM, Move, WASM, and ZK interfaces.
  • Rohu-Paplet: Byzantine consensus research, including the relationship between ordering, execution, observation, and finality.
  • Correctness and security: EVM differential testing, failure testing, and adversarial protocol analysis.
  • Market mechanisms: DRI and Asset Factory work on asset issuance and index tracking, and Orbital experiments in AMM design. The DRI paper documents the earlier design.
  • Interfaces and integration: command-line and MCP tooling, plus work connecting execution and consensus to persistent state, including QMDB integration.

Selected engineering and research

Area Work
GPU systems
CUDA Rust
Inference kernels for mixture-of-experts models and decode workloads. FlashMap is a bulk hash map in Rust with CUDA acceleration and CPU/Rayon fallbacks.
Compilers and execution
Vx EVM
EVM-vx, a research-stage CPU EVM with optimistic parallel execution; vex, a prototype package manager; and upstream Vx compiler contributions.
Custody and authorization
Rust Solana
Open Institutional Custody: Rust components for payment approvals, spending limits, and EVM/Solana signing. A development preview with simulated chains.
Verifier-guided learning
Live demo
Daybreak: experiments in independent emulator checks, bug-finding workflows, and verifier-generated training data. Interactive demo.
Strategic reasoning
Python Rust
Sigma-Go: research on counterfactual value prediction, poker AI, and learning in imperfect-information games.
Agents and open networks
MCP TypeScript
NANDA server and plugin integration, built on the MCP TypeScript SDK, and Beckn AI tooling for developer assistance and retrieval.

Work in practice

  • Exelon / PECO: GasGPT and production AI infrastructure for gas operations, including backend modernization supporting 1,000+ users.
  • Beckn Protocol: an AI internship working on transaction interfaces and a knowledge-based developer assistant.
  • Drexel: Computer Science, research, and teaching support.
Earlier ventures and applied ML
  • Encryptech / Mycelium: decentralized and privacy-preserving machine learning, including federated learning and FHE. The public architecture notes describe the federated-learning and Beckn direction.
  • Classcut / MPX: educational software, video search, and ML product work.
  • Dynamic Rocketry: earlier aerospace work on solid rocket motors and cold-gas thrusters.
  • Delphic: API-to-chain oracle experiments, with contracts, off-chain ingestion, and TEE-oriented components.
  • Indic LLM Toolkit: data preparation, training, evaluation, and inference experiments for multilingual models.
  • Insurance ML: experiments combining tabular, graph, text, and anomaly-detection models.
Developer tools and technical explorations
  • DeltaMCP: git-delta integration for readable diffs and Git views in coding assistants.
  • STE writing: technical-writing tools with agent edit hooks and Vale/Harper linting.
  • Jev / Codex integration: a Codex history adapter and post-compaction context restoration contributed in my fork.
  • Research forks: Solidus, for Lean 4 and Yul-to-EVM compilation, and strevm, for streaming asynchronous EVM execution.

Milestones

HackMIT Codebase track: 2nd place ICPC North America Qualifier: 2nd place at Drexel
$50K foundation grant Provisional patent filing for dynamic market-maker architecture

Tools

Rust CUDA / PTX Go C / C++
Python TypeScript Solidity Lean 4

Pinned Loading

  1. VYRN VYRN Public

    xchng

    Rust 1

  2. solve solve Public

    solve

    HTML 3 2

  3. EVM-vx EVM-vx Public

    Rust 2

  4. indic-llm-toolkit indic-llm-toolkit Public

    HyperSpecialised LLM optimisation on regional data

    Python 1

  5. Nanda-Sdk-Tsc Nanda-Sdk-Tsc Public

    TypeScript 3 1

  6. delphic delphic Public

    TypeScript 1

0