Tony Joseph Sebastian| Developer

About Me

🤖 AI & ML-focused Software Engineer with over 5 years of experience in designing and deploying machine learning and GenAI solutions.

☁ Proven ability to build scalable cloud-native ML applications, conduct end-to-end model training and evaluation, and develop intelligent automation tools.

🛠️ Extensive hands-on experience with LangChain, Azure AI, OpenAI, and FastAPI.

⚙ Adept in CI/CD practices, unstructured data processing, and integrating ML pipelines into enterprise cloud infrastructure.

✨ Seeking to bring innovative GenAI and ML expertise into transformative enterprise solutions.

Work Experience

TD Bank Group

Software Engineer I (Full-Time)

Jan 2023 – Present | Toronto, Canada

Software Engineer Intern

Sept 2022 – Dec 2022

Built Generative AI-powered platforms using LangChain and Azure OpenAI to support internal legal, financial, and compliance automation tools.
Designed backend APIs using FastAPI for real-time interaction between LLMs and internal TD data lakes.
Developed and maintained AI-enabled tools:
- TD Lawyer Me: Legal document analysis and suggestion platform that scans vendor contracts, compares them with TD's internal policies, and generates deviation reports and document drafts. Technologies used: React, FastAPI, Azure AI Search, Azure OpenAI, LangChain, LangGraph, GitHub Workflows.
- TD Codex: Intelligent code analysis and generation platform for legacy and modern languages such as COBOL, SAS, SQL, and SQR. Uses LLMs to parse, refactor, and document large codebases.
- TD Document Chat AI: Built on LangChain and FAISS, it allows employees to query TD's internal documents and policies in a conversational format using Azure OpenAI.
- FRP Mapping Tool & OFSAA Mapper: Generative AI-powered internal finance tools that automate reconciliation and metadata mapping. Built with Streamlit, Docker, and Azure OpenAI, deployed on Windows Server environments.
Trained, evaluated, and retrained models to reduce drift; integrated model monitoring and reporting.
Built evaluation frameworks and automation tests to validate GenAI results.
Set up CI/CD pipelines and containerized GenAI services for Windows Server environments.
Applied ML to unstructured data (e.g., contracts, policy docs) using embeddings, text classification, and tagging.

Technologies: Python, FastAPI, React, LangChain, Azure AI, OpenAI, PostgreSQL, MongoDB, Docker, CI/CD.

Hewlett Packard Enterprise (via Inspirisys Solutions)

Software Engineer (C++ / Visual C++)

June 2018 – July 2021 | Trivandrum, India

Led Proof-of-Concept (PoC) development for AI-enhanced features in enterprise scanning tools, improving automation and document classification workflows. Prototypes presented to stakeholders for validation, increasing customer adoption and satisfaction.
Built and optimized image processing pipelines using C++, OpenMP, and Chronos libraries, resulting in a 20% boost in processing speed for HP Scan applications.
Implemented custom C++ device drivers using WIA and Tulip frameworks, integrating HP hardware with AI-driven scanning solutions.
Developed intelligent cloud upload plugins (C++/CLI) to Box, Dropbox, Google Drive, etc., enabling seamless document transfer for further machine learning processing and OCR tasks.
Designed modular C++ algorithms to support image quality detection, classification, and enhancement—laying the groundwork for future AI/ML integrations.
Built unit and integration test suites using Google Test and CPPTest, and deployed CI/CD pipelines with Jenkins to automate testing and reduce regression errors.
Collaborated in Agile/Scrum environment with global teams; participated in sprint reviews, defect triage, and best practice discussions.
Created technical documents, conducted internal demos, and supported customer success teams with AI-enhanced troubleshooting tools.
Integrated early-stage Python-based ML models into internal test environments for experimentation with document classification and metadata extraction.

Projects: HP Scan Plus, HP Universal Scan, HP Twain Plus, HP Ink Supply Station, HP Scan Cloud. Technologies: C++, Python, OpenMP, MFC, WPF, .NET, WIA, Tulip, Node.js, Jenkins, SVN/Git, Google Test, CPPTest, HP Cloud, Dropbox.