Mustahid Ahmed

IBM System Engineering

March 2025 - Present

Pioneer Electronics

July 2023 - Feb 2025

Sensyn Robotics

May 2022 - June 2023

HyperCube Ltd.

April 2020 - April 2022

Batton Ltd.

Aug 2023 - Dec 2023

AI Idea Lab Ltd.

April 2023 - Dec 2023

Open Source Contributor

April 2020 - Present

Financial AI Multi-Agent System,IBM System Engineering

April 2025 - Present

Developed a sophisticated multi-agent AI system for financial data analysis and document processing using LangGraph orchestration framework. Built a scalable FastAPI backend integrating specialized AI agents for SQL query execution, PDF document retrieval, data visualization, and intelligent request routing. Implemented agentic capabilities with multilingual support (English/Japanese) for financial document analysis.

Key Contributions:

• Architected state-based agent workflow system using LangGraph for coordinated AI agent interactions
• Built RESTful API endpoints with session management and CORS support for web application integration
• Implemented automated financial data visualization and chart generation using Matplotlib
• Developed PDF processing pipeline for extracting and analyzing financial documents
• Set up comprehensive development environment with code quality tools (Ruff, Basedpyright)

PostgreSQL MCP Server: Dual Transport & Database Integration,Open Source Contributor

January 2025

Developed a production-ready PostgreSQL MCP (Model Context Protocol) server with dual transport support (HTTP and Stdio). Featured in MCP official repository.

Key Contributions

• Architected a dual-transport system (HTTP & Stdio) and designed a RESTful API using FastAPI, complete with automatic OpenAPI documentation.
• Engineered a robust PostgreSQL integration featuring connection pooling, stateful session management, and secure, parameterized queries to prevent SQL injection.
• Developed a flexible configuration system using environment variables and command-line arguments to enhance deployment adaptability.
• Streamlined deployment by creating production-ready Docker containers and Docker Compose configurations.
• Implemented comprehensive error handling, graceful shutdown logic, and a full testing suite to ensure system reliability and stability.

xiyan_mcp_server: Azure OpenAI, llama-cpp-python, & SQLite Integration,Open Source Contributor

January 2025

Contributed to xiyan mcp server project (Natural Language to SQL MCP server) by implementing comprehensive Azure OpenAI integration and enhancing local model support through llama-cpp-python. Enabled Streamable HTTP transport protocol and integrated SQLite database support.

Key contributions:

• Implemented Azure OpenAI SDK integration with configurable model endpoints and API versions
• Added Streamable HTTP transport support with configurable host/port parameters
• Enabled llama-cpp-python local model support for running models locally
• Integrated SQLite database backend with configurable dialect and database path
• Enhanced configuration system with comprehensive YAML-based setup
• Added comprehensive error handling and logging capabilities
• Created detailed documentation and usage examples for all new features

Implemented mcp server for static error correction with basedpyright,Open Source Contributor

May 2025 - Present

Contributed to quack-mcp-server project by adding BasedPyright static analysis support as an MCP tool alongside existing mypy integration. BasedPyright provides faster type checking with enhanced features compared to standard pyright.

Key contributions:

• Implemented automatic installation and configuration detection for basedpyright
• Added severity filtering and top-N limiting capabilities for diagnostic output
• Created comprehensive test suite with full coverage for edge cases and error handling
• Integrated with existing diagnostic filtering utilities for consistent output format

Implemented RAG pipeline with LLM for Qualcomm SoC,Pioneer Electronics

April 2024 - August 2024

Designed and implemented end to end RAG pipeline from scratch using faiss and llama.cpp.

• Created separate library for RAG after deriving llama and embeddings from llama.cpp
• Implemented build system using Conan
• Set up CI/CD using GitHub Actions

Designed end to end middleware system for automotive radar,Pioneer Electronics

August 2023 - March 2024

Implemented API that obtains sensor data streams and transfers them to machine learning module to deliver inference results back into an alarm system.

• Implemented lock-free ring buffer system that stopped data loss and enabled extremely fast inference
• Employed comprehensive unit tests and CI/CD pipeline
• Delivered real-time inference results to alarm system

Crack segmentation from tiles,Sensyn Robotics

Feb 2023 - April 2023

Implemented model for crack segmentation using novel transformer-based segmentation architecture inspired from retinal vessel segmentation in medical literature.

• Implemented continual learning system based on labelling→training cycle
• Achieved high precision of 0.9 and recall of 0.8
• Applied cutting-edge medical imaging techniques to civil engineering problems

Analog gauge reading by computer vision,Sensyn Robotics

November 2022 - April 2023

Constructed model for accurately reading analog gauges by determining dial position, minimum and maximum values from the scale face after applying perspective transformation.

• Minimized expected error rate to 0.15%
• Compressed model size through quantization in C++ using TVM
• Successfully deployed to edge device (NVIDIA Jetson Nano)
• Applied advanced computer vision techniques for precise measurement

People detection from fish eye images,Sensyn Robotics

May 2022 - July 2022

Developed new algorithm for people detection in fisheye images, as commonly used algorithms are not suitable for such distorted perspectives.

• Incorporated angular values of slanted bounding boxes for improved detection accuracy
• Implemented end-to-end pipeline from labelling to cloud storage using CoCo Annotator and AWS S3
• Achieved high mAP score comparable to state-of-the-art YOLO algorithms
• Solved unique challenges posed by fisheye lens distortion

NLP based Chatbot, HyperCube Ltd.

June 2021 - October 2021

Full stack Development: Frontend and backend development using Flutter/Dart/Swift. Model deployment, system design and maintenance using AWS services:

• AWS Amplify for frontend hosting
• AWS API Gateway for API management
• AWS Lambda for serverless functions
• AWS SageMaker for ML model deployment
• Frontend and backend development using Flutter/Dart/Swift

Database design and deployment, HyperCube Ltd.

March 2021 – May 2021

Designed MySQL database schema and deployed it in Azure.

• Created comprehensive database schema design
• Deployed database on Azure cloud platform
• Developed web application API for database connectivity
• Implemented secure data access patterns

Designed novel algorithm for OCR of Japanese fax documents,Batton Ltd.

Aug 2023 - Dec 2023

Collected and labelled dataset of fax documents containing tables. Implemented novel algorithm to extract its data and obtained 97% accuracy.

• Collected and labeled dataset of fax documents with tables
• Developed novel OCR algorithm for table data extraction
• Achieved 97% accuracy on test dataset
• Dockerized the ML model for deployment
• Created HTTP endpoint using TorchServe
• Deployed to GCP Artifact Registry
• Created inference endpoint through Vertex AI

Japanese kanji generation from a specific font,AI Idea Lab Ltd.

April 2023 - Sep 2023

Used controlnet and prompt engineering to create Kanjis from novel fonts.

• Applied ControlNet for Japanese kanji generation
• Used prompt engineering for font-specific character creation
• Built interactive frontend using Gradio
• Developed FastAPI backend endpoints
• Integrated Gradio frontend with FastAPI backend

GPT based AI Callerbot,AI Idea Lab Ltd.

Oct 2023 - Dec 2023

Converting caller's speech into text and using prompt engineering to obtain the best output from Open AI API.

• Convert caller's speech to text using speech recognition
• Apply prompt engineering for optimal OpenAI API responses
• Convert API responses back to voice using text-to-speech
• Create WebRTC bridge between Python and Node.js
• System currently being deployed in company application

Memorize: A game for training memory,Open Source Contributor

June 2021 - Sep 2021

Designed memory training game with engaging user interface.

• Designed frontend using Swift programming language
• Implemented game logic for memory training mechanics
• Created intuitive user experience for cognitive exercises

amusta-chain: A distributed peer to peer blockchain,Open Source Contributor

June 2021 - Sep 2021

Developed distributed peer-to-peer blockchain system with comprehensive testing.

• Employed end-to-end Test-Driven Development (TDD)
• Implemented peer-to-peer mining functionality
• Created secure currency transfer system to user wallets
• Developed transaction validation by miners
• Built distributed consensus mechanism