Vaibhav Goswami

Vaibhav Goswami

Leading the Generative AI Platform at Thomson Reuters. Orchestrating multi-LLM systems for 20,000+ users.

About Me

I architect and build AI systems that democratize access to cutting-edge technology at enterprise scale.

The platform I lead unifies access to frontier models — Claude, GPT, Gemini, and Llama — alongside custom fine-tuned models deployed on SageMaker, enabling non-technical professionals to build sophisticated AI workflows without writing a single line of code. This work was recently featured on AWS's Machine Learning Blog as a case study in enterprise AI architecture.

Technical Focus

  • ▸ Python-first architecture and design
  • ▸ LLM integration & multi-provider orchestration
  • ▸ RAG (Retrieval-Augmented Generation) at scale
  • ▸ Serverless architectures (AWS Lambda, DynamoDB)
  • ▸ ML monitoring & observability

I'm passionate about making AI accessible, building systems that scale, and sharing learnings with the community. Always happy to discuss AI architecture, Python engineering, or the challenges of production ML systems.

Technical Expertise

Building enterprise-scale AI platforms with modern cloud infrastructure and cutting-edge technologies

AI & ML

Generative AI Expert
LangChain Expert
RAG Systems Expert
MLOps Expert
TensorFlow Proficient
Computer Vision Proficient

Cloud & DevOps

Google Cloud Expert
AWS Expert
Kubernetes Proficient
Terraform Proficient
Docker Proficient
CI/CD Proficient

Development

Python Expert
FastAPI Expert
SQL Proficient
REST APIs Expert
Git Expert
Microservices Proficient

Additional Technologies & Tools

Keras Flask Scikit-learn OpenCV OpenAI SDK AWS Bedrock Vertex AI SDK DynamoDB BigQuery Snowflake SageMaker Vector Databases

Professional Journey

Senior AI Software Engineer

Thomson Reuters

June 2022 - Present
  • ● Architected and implemented the core serverless infrastructure serving 20k users, designing the LLM provider abstraction layer and workflow orchestration engine. Mentored 10+ engineers on the technical implementation.
  • ● Built capabilities for users to create sophisticated Agentic AI solutions, enabling them to chain multiple databases and agents together to solve complex business problems.
  • ● Developed an automated production-grade RAG deployment system from scratch, allowing users to deploy retrieval-augmented generation solutions with enterprise-level reliability and scalability.
  • ● Architecting a comprehensive platform that democratizes AI capabilities, providing self-service tools for model development, deployment pipelines, and production monitoring.
  • ● Co-authored the AWS Machine Learning Blog post Democratizing AI: How Thomson Reuters Open Arena supports no-code AI for every professional with Amazon Bedrock.

Senior Software Engineer

HSBC Technology

Sept 2021 - May 2022
  • ● Previously built ML platforms reducing model deployment time by 90% through Kubernetes automation.
  • ● Led cloud migration initiatives, managing the transition of data from on-premises to Google Cloud Platform.
  • ● Implemented automated Terraform pipelines, streamlining production deployments and reducing errors by 95%.
  • ● Developed Infrastructure as Code (IaC) for Compute Engines, Managed Instance Groups, Load Balancers, Cloud Storage Buckets, and BigQuery resources.

Software Engineer

HSBC Technology

July 2019 - Aug 2021
  • ● Developed an Automated Machine Learning (AutoML) Pipeline, automating 90% of the manual work in training and deploying models.
  • ● Designed and deployed a Python-based Automatic Test Data Generator for testing Dev and UAT pipelines, utilizing the Dash-Flask framework.
  • ● Created a Python Web API hosted on Google Cloud, enabling search functionality for over 50+ data assets within the organization.

Machine Learning Intern

UST Global

July 2018 - Dec 2018
  • ● Developed a Car Damage Detection Algorithm that estimates damage on individual car parts from images.
  • ● Created a Mask-RCNN model for object detection, segmenting car parts with 91% accuracy.
  • ● Built a multi-class image classification model to identify the severity of car damage (Minor, Moderate, Severe), achieving 83% accuracy in identifying damaged parts.

Projects

CodeGraph

Interactive Code Dependency Visualizer

A beautiful, interactive web application that visualizes GitHub repository dependencies as stunning force-directed network graphs. Drag nodes, click to highlight connections, and explore code architecture like never before.

Key Features

● 🕸️ Interactive Force Graph: Drag, zoom, and pan to explore large codebases with smooth physics simulations.
● 📊 Rich Metrics: Cyclomatic complexity, lines of code, and dependency tracking per file.
● 🎨 Visual Encoding: Blue gradient complexity scale (light blue = simple → deep blue = complex), node size reflects lines of code.
● 🔍 Smart Highlighting: Click nodes to highlight dependencies and understand module coupling.

Tech Stack

Python FastAPI React TypeScript D3.js NetworkX Docker

Snap Sudoku Solver

AI-Powered Web Application

An AI-powered web application that solves Sudoku puzzles from camera images in seconds. It uses computer vision to detect the grid and deep learning OCR to recognize digits.

Key Features

● 📸 Instant Digitization: Auto-crops and rectifies puzzle grids from photos.
● 🧠 Smart OCR Pipeline: Custom-tuned OCR with noise filtering.
● ✍️ Handwritten Support: Recognizes handwritten digits, perfect for validating partially solved puzzles.
● 🎨 Interactive UI: Users can manually correct digits before solving.

Tech Stack

Python Flask OpenCV Deep Learning OCR

Let's Connect

Open to connecting with folks working on AI infrastructure, MLOps, and production LLM systems.

Download Resume