Skip to content
View Manishmaurya89's full-sized avatar
🎯
Focusing
🎯
Focusing
1 follower · 2 following

Achievements

Achievement: Pull SharkAchievement: YOLO

Achievements

Achievement: Pull SharkAchievement: YOLO

Block or report Manishmaurya89

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
Manishmaurya89/README.md

Manish Maurya

I build LLM systems and then try to break them: fine-tuning for Indic languages, synthetic training data, and security for AI agents that use tools.

Research Intern @ LTRC, IIIT Hyderabad  ·  Portfolio  ·  LinkedIn  ·  Email


MCP-GuardBench dashboard

MCP-GuardBench

Is the MCP server behind your AI assistant poisoned? GuardBench checks, and benchmarks security tools against 11 deliberately vulnerable servers. My runtime guard blocks 8 of 9 attacks with zero false alarms.

pip install mcp-guardbench 		Synthetic Data Generator output

Synthetic Data Generator

Training sentences in 13 Indian and South Asian languages from any local LLM, with no API key. Script checks, dedupe and exact language mixing keep the data clean. Gemma 4 wrote 27/27 valid Urdu sentences; Llama 3 managed 9.

Also built

  • repo-bug-hunter: an AI agent that fixes real GitHub bugs, graded by hidden tests, with every run replayable. pip install repo-bug-hunter
  • Synthetic Document Factory: fake documents, real labels. PDFs, page images and exact COCO/YOLO boxes for layout detection.
  • Basketball Analytics: player and ball tracking, zero-shot team split, passes and possession on a top-down court. Extended from an open-source project.

PyTorch Hugging Face LoRA / QLoRA RAG NER Ollama FastAPI Docker

Open to AI/ML, GenAI and LLM engineering roles in Bengaluru, Hyderabad or remote.

Pinned Loading

  1. Basketball-Analytics-Estimation-Using-Object-Segmentation- Basketball-Analytics-Estimation-Using-Object-Segmentation- Public

    Basketball game analysis tool using deep learning for object detection, player tracking, and team assignment.

    Jupyter Notebook 1

  2. Synthetic-data-generator- Synthetic-data-generator- Public

    Generate free synthetic training data in Indian & South Asian languages with any local LLM via Ollama. No API keys, no token limits.

    Python

  3. Face-direction-detection Face-direction-detection Public

    Real-time face direction detection in the browser with TensorFlow.js and MediaPipe FaceMesh

    JavaScript

  4. MUSIC-RECOMMENDATION-USING-FACIAL-RECOGNITION MUSIC-RECOMMENDATION-USING-FACIAL-RECOGNITION Public

    Music Recommendation using Facial Expression

    Python

  5. mcp-guardbench mcp-guardbench Public

    Check your MCP servers for tool poisoning and rug pulls, and benchmark MCP security tools under verifiable conditions.

    Python 1

  6. repo-bug-hunter repo-bug-hunter Public

    An AI coding agent that fixes real GitHub bugs, and a lab that measures how well it does it: graded by hidden tests, every run replayable.

    Python