About Me

A Senior Data Scientist & AI/ML Engineer with 5+ years of experience building production-grade AI systems that deliver real business impact. My current focus is on designing and maturing AI agents, and agentic workflows, moving them from experimental prototypes to reliable, autonomous systems that operate at scale.

• My core expertise spans Agentic AI Systems, LLMs, Generative AI, MLOps, and Data Engineering, with deep hands-on experience across Healthcare, Legal, Finance, Logistics, Sales, and Customer Support.

• I've architected solutions for industry leaders including Motorola, Caliber Home Loans, and Logitech, where the measure of success was always business outcome, not model accuracy alone.

• I operate at the intersection of engineering and strategy, leading AI/ML teams, owning end-to-end pipelines from data to deployment, and working directly with stakeholders to turn complex AI capabilities into decisions that move the needle.

Outside of work, I follow AI research and industry trends closely and enjoy exploring new food spots, because great ideas, like great food, are worth seeking out.

If you're building something ambitious with AI, let's talk.

Skills

Experience

Education

Certifications

  • Areas of Expertise
    • Machine Learning & AI | Conversational AI & NLP Solutions | MLOps & Model Deployment | Data Modeling & Feature Engineering | GenAI & FinTech Solutions
  • Languages
    • Python, SQL, C/C++, Java, Bash
  • Machine Learning
    • Classification, Regression, t-SNE, PCA, CNN, UNet, OpenCV, NLTK, SpaCy
  • Frameworks
    • Sciki-learn, TensorFlow, PyTorch, Keras, SciPy, XGBoost, Hyperopt, Pandas, NumPy, Vaex, Dask, PySpark, MLflow, Streamlit
  • Data Science
    • Statistical Analysis, Data Mining, Data Wrangling, Data Drift, Concept Drift, A/B Testing, T-Test
  • Generative AI
    • OpenAI, Ollama, Deepseek, Cohere, Mistral, Gemini, Microsoft Phi, Vertax, LangChain, LlamaIndex, HuggingFace, Transformers, RAG (Naive, Rerank, Agentic), LangGraph, NL2SQL, Vector Database (Chroma, Qdrant, Faiss, Pinecone), PEFT, LoRA, DSPy, Explainable AI, Prompt Engineering, Bias Mitigation, Ragas
  • Database Management
    • MySQL, SQL Server, PostgreSQL, MongoDB, Snowflake, Kafka, Salesforce SOQL, Oracle, Elasticsearch, MariaDB, NoSQL
  • Data Visulization
    • Power BI, Matplotlib, Seaborn, Plotly, Dash
  • Cloud & DevOps
    • AWS, Azure, GCP, Sagemaker, Bedrock, Databricks, Hadoop, AWS Lambda, Serverless, Docker, Kubernetes, Airflow, Git, Jenkins, SSH, Github Actions, Infrastructure-as-Code
  • Mar 2022 - Present (Giggso Inc.)
    Data Scientist
  • Dec 2020 - Mar 2022 (Giggso Inc.)
    Junior Data Scientist
  • Aug 2018 - Feb 2021 (Fast NUCES University)
    Teaching Assistant
  • Oct 2019 - Sep 2020 (Fast NUCES University)
    Full Stack Intern
  • National University of Computer & Emerging Sciences (NUCES)
    Bachelor's of Computer Science
  • Government Graduate College
    Intermediate in Computer Science
  • Anthropic
    Claude Code in Action
  • Oracle
    OCI Certified Data Science Professional
  • Databricks
    Advanced Machine Learning Operations
  • Microsoft
    Microsoft Certified: Azure AI Fundamentals
  • Databricks
    Academy Accreditation - Generative AI Fundamentals

My Services

AI & Machine Learning Solutions

Design and deploy intelligent systems for predictive analytics.

Learn more

Generative AI & NLP Services

Create AI-driven content and advanced language models.

Learn more

MLOps & AI Infrastructure

Streamline model deployment and automate workflows.

Learn more

Data Engineering & Business Intelligence

Build robust data pipelines and interactive dashboards.

Learn more

AI Strategy & Consulting

Align AI initiatives with your business goals.

Learn more

Web Scraping & Automation

Extract and process web data at scale.

Learn more

My Work

Enterprise AI Work at Giggso Inc.

Production AI systems built for enterprise workflows, governance, recruiting, CRM, conversational intelligence, and model operations.

Mar 2026 - Present

Raven Enterprise

Enterprise

AI coding governance and guardrails platform for secure repo-aware task allocation, LLM spend visibility, and compliance monitoring.

  • Mapped AI-assisted activity across 50+ repositories to Git commits with 98% attribution accuracy.
  • Built automated risk controls for red-teaming, secret detection, IAM scoring, and pre-commit data leak prevention.
  • Designed telemetry-driven LLM routing that reduced token costs by 32%.
Claude Code MCP LLM Governance PostgreSQL Observability
Nov 2025 - Mar 2026

Reveilio

Enterprise

AI recruitment platform for resume and job description parsing, candidate shortlisting, and assessment report generation.

  • Built async resume ingestion with embeddings and ChromaDB, processing multi-format documents in under 2 seconds.
  • Developed a hybrid RAG and BM25 matching engine that improved candidate relevance by 28%.
Python OpenAI FastAPI ChromaDB Docker
Mar 2025 - Oct 2025

Sales Copilot

Enterprise

Conversational Salesforce assistant that lets sales teams query CRM data in natural language without writing SOQL.

  • Built schema-filtered RAG for natural language CRM queries with a 91% valid SOQL execution rate.
  • Created a semantic query cache that cut token spend by 45% and reduced repeat-query latency to 180ms.
OpenAI SOQL RAG FastAPI LangChain
Feb 2024 - Mar 2025

Shay

Enterprise

Conversational intelligence platform for sales and support teams with secure document understanding and semantic retrieval.

  • Built key-value, tabular, PDF text, table, and image extraction pipelines for context-aware search.
  • Integrated Cohere and Gemini through OCI Generative AI and GCP with field-level authorization.
Python Gemini Cohere RAG Vector Search
May 2023 - Jan 2024

Trinity ModelOps

Enterprise

Production ML observability platform for model monitoring, data quality checks, drift detection, and retraining alerts.

  • Implemented validation pipelines for schema, range, null-rate, and feature integrity monitoring.
  • Built KS-test and PSI-based drift detection with alerting for retraining workflows.
FastAPI Pandas Sklearn MLOps Monitoring
Apr 2021 - May 2023

Earlier Giggso AI/ML Systems

Enterprise

Log analytics, anomaly detection, MCP-integrated tooling, and NLP intent classification systems for enterprise operations.

  • Converted log analysis capabilities into MCP tools callable by AI assistants.
  • Built template-free deep learning anomaly detection and BERT-based intent classification workflows.
MCP BERT TensorFlow Elasticsearch AWS

Generative AI Projects

Mermaid Health

Transforming Healthcare with AI - Simple, Seamless, and Personalized

An AI-powered solution transforming clinical assessment report writing and health management. Mermaid Health automates the generation of comprehensive clinical reports, reducing processing time while enhancing diagnostic accuracy. By leveraging advanced AI models, it streamlines documentation workflows, enabling healthcare professionals to focus on patient care rather than administrative tasks.

Skills: Tensorflow, NLTK, SpaCy, LangChain, LangGraph, Chroma, Docker, OpenAI, Fitz

Factual Inquirer

An AI-Powered Platform for News Credibility & Censorship Assessment

An AI-driven platform that evaluates news credibility and censorship transparency, providing both free and premium insights. Fine-tuned models for News Value (NV) and News Censorship Transparency (NCT) to generate structured JSON outputs with numerical scores and actionable improvement suggestions. Developed a Credibility Named Entity Recognition (CNER) system using OpenAI, leveraging rule-based techniques to rank articles based on credibility parameters. Additionally, built ChatBox, a journalist-focused web application that streamlines article submissions, profile management, peer collaboration, and customer support.

Skills: Transformers, Scikit-Learn, Tensorflow, PyTorch, Keras, HuggingFace, LoRa, PEFT

BidMaster AI

AI-Powered RFP Processing & Proposal Automation

Designed and implemented an AI-driven solution for RFP and RFI processing, automating requirement extraction, classification, and analysis. Developed an intelligent Proposal and Bidding Document Generation system to create structured, high-quality responses, reducing manual effort and ensuring compliance. Leveraging NLP-powered automation, the system enhances accuracy, efficiency, and consistency in the proposal and bidding process.

Skills: NLTK, SpaCy, PyTorch, LangChain, LangGraph, Azure Openai, Fitz, PDFPlumber, Docx

Natural Language Query AI

Conversational AI for Seamless SQL & SOQL Generation

Developed an AI-powered system that enables users to interact with their databases and Salesforce objects using natural language. Implemented query classification to intelligently route user queries to the relevant data sources. Designed and optimized SQL and SOQL generation pipelines using a custom data source schema and RAG models, ensuring efficient and accurate query formation. This solution enhances data accessibility, allowing users to retrieve insights effortlessly without writing complex queries.

Skills: Pandas, LangChain, LangGraph, RAG, Azure Openai, Salesforce APIs

Data Science Projects

Product Demand Forecasting

ML driven time-series product demand forecasting to optimize inventory and business planning.

Developed a machine learning-powered time-series forecasting solution to predict product demand, enabling businesses to optimize inventory management and strategic planning. Implemented advanced ML models to analyze historical sales data, identify trends, and generate accurate demand forecasts. This solution helps businesses reduce stockouts, minimize overstocking, and enhance supply chain efficiency, leading to improved decision-making and cost savings.

Skills: Pandas, NumPy, Sklearn, Matplotlib, XGBoost, HyperOpt

PhishGuard

Real-time phishing URL detection system using Machine Learning and PhishTank database verification.

Built a machine learning-powered phishing detection system that analyzes URLs in real-time to identify potential phishing attempts. A Random Forest classifier trained on 20,000+ URLs achieves 99% accuracy by extracting 60 URL features including entropy, domain analysis, and suspicious patterns. The system integrates with the PhishTank database for cross-referencing against known phishing URLs and provides a FastAPI-based REST API for instant verification alongside an interactive Bootstrap 5 web dashboard with statistics, alerts, and severity levels.

Skills: Python, Scikit-learn, FastAPI, Random Forest, Pandas, NumPy, Bootstrap, PhishTank API

Nuclei Spotter

Detected infected region of Nuclei from divergent images with UNet

Developed a deep learning solution using the UNet model to accurately detect and segment nuclei in diverse microscopic images. Leveraging advanced image processing techniques, the system identifies nuclei across varying conditions, enhancing biomedical research and diagnostic accuracy. This project improves automated cell analysis, supporting applications in medical imaging, pathology, and drug discovery.

Skills: NumPy, Skimage, Matplotlib, TensorFlow

Intent Recognition

Intent Recognition Using BERT on Colab TPU

Developed an Intent Recognition Engine leveraging the BERT (Bidirectional Encoder Representations from Transformers) model to accurately classify user intents from natural language input. Implemented the solution using TensorFlow on Google Colab TPU to accelerate training and inference. BERT's deep contextual understanding pretrained on Wikipedia and large corpora enabled high-performance intent detection across diverse inputs. This project showcases the application of state-of-the-art transformer-based NLP models for real-time language understanding tasks.

Skills: NumPy, Pandas, TensorFlow, Bert, Sklearn, Matplotlib, Seaborn

Recommender System

ML-powered suggestions to resolve application errors

This recommender system is designed to assist in resolving application issues by analyzing error logs and providing context-specific solutions. When an error log is generated, it is first classified as normal or anomalous. If identified as an critical anomaly, the system searches across business documentation, historical group discussions, and pre-configured resolution data stored in Elasticsearch to suggest the most relevant fixes. It returns the top 5 recommended solutions, each with a confidence score, enabling users to take quick and informed action to resolve critical errors efficiently.

Skills: NumPy, Pandas, TensorFlow, BERT, Sklearn, Elasticsearch

Data Engineering & ETL

AnomalyFlow

Scalable Log Pipeline for Smart Anomaly Routing

AnomalyFlow is a scalable data engineering pipeline built to streamline log ingestion and anomaly detection. It ingests data from multiple client servers into Kafka, processes it through an anomaly classification engine, and routes the results accordingly publishing anomalies to hot storage (Kafka and Elasticsearch) for real-time analysis, while directing non-anomalous data to cold storage for long-term retention. Designed to enhance analytical capabilities, AnomalyFlow brings structure, speed, and intelligence to large-scale log processing systems.

Skills: Pandas, SpaCy, NLTK, Sklearn, Kafka, Elasticseach, Kibana

Invoice Processing

Automated Invoice Processing and KPI Reporting

Developed end-to-end ETL pipeline designed to extract and standardize payment receipt data from Citrix ShareFile storage. Each customer folder contains a variety of receipt formats including Images, PDFs, CSVs, and Excel files. The pipeline reads and processes these receipts using OCR for image files and PDF/data parsing tools for other formats. Extracted information is transformed into a standard JSON structure, then loaded into a MySQL database for querying. Users can perform aggregation operations, track business KPIs, and view dynamic reports based on the structured data, enabling smarter financial insights from previously unstructured content.

Skills: Pandas, SpaCy, OCR, PyPDF2, MySQL, Seaborn

LegalMine

Large-Scale Legal Data Extraction, Processing & Vector Search Pipeline

Architected a large-scale ETL pipeline to extract millions of articles, judgements, case laws, and legislations from diverse legal sources across multiple jurisdictions. Built custom scraping solutions per source using Scrapy, BeautifulSoup, Playwright, Jina AI, and Firecrawl with site-specific logic for pagination, PDF extraction, and dynamic content. Leveraged Bright Data Proxy to bypass Cloudflare and server-side blocking.
Preprocessed and normalized data was persisted in MongoDB, then embedded and indexed in Weaviate to enable semantic search, vector similarity with keyword filtering, and RAG for intelligent legal research.

Skills: Scrapy, BeautifulSoup, Playwright, Jina AI, Firecrawl, Bright Data, MongoDB, Weaviate, Python, httpx

Contact Me

mahsan12030@gmail.com

Visit Blog

Copyright © All rights reserved.