Available for Opportunities

GhanshyamSoni

Software that ships. Systems that learn.

Full Stack  ·  Applied AI & ML  ·  Cloud  ·  5+ Years

View My WorkDownload Resume
Toronto, ON · Canada
About

I build software that works — and lately, software that learns.

5+Years Experience
Top 5AWS DeepRacer Finish
10+Production Systems
CurrentlySr Full Stack Developer · SFI Group Ltd · Toronto

I'm a software engineer who has spent five years shipping production systems — operations platforms, data pipelines, manufacturing workflows — across three companies and two continents. Things that real teams depend on every day, not demos.

Then I trained a reinforcement learning model for an AWS DeepRacer competition, watched it teach itself to take corners, finished top 5, and got hooked. Since then I've been pulling machine learning into everything I build — LLM-powered features, retrieval pipelines, model training on SageMaker — on top of the full-stack engineering that pays the bills.

I care about systems that hold up under real load, code the next person can maintain, and models that do something useful instead of just impressive.

Based in Toronto. Open to software engineering and AI/ML engineering roles — remote-ready, fast learner by track record.

PythonTypeScriptReactLaravelNode.jsAWS · SageMakerAzureDocker
LinkedInGitHub
Technical Skills

The Stack

5+ years across full-stack engineering, cloud, and — increasingly — applied AI. Highlighted = primary expertise.

AI & Machine LearningAdvanced
PythonReinforcement LearningAWS SageMakerLLM APIs (Claude, OpenAI)RAG PipelinesPrompt EngineeringVector Search (pgvector)Pandas / NumPyscikit-learnModel Evaluation
BackendExpert
Java 8–17Spring BootSpring MVCPHP (Laravel)Node.jsC# / ASP.NETREST APIsJPA / HibernateMicroservicesModular Monolith
FrontendAdvanced
React.jsReact NativeVue.jsTypeScriptJavaScript ES6+HTML5 / CSS3BootstrapSPA Architecture
Cloud & DevOpsAdvanced
AWS (EC2, S3, Lambda)Microsoft AzureDockerKubernetesCI/CD PipelinesTerraformJenkinsGitLab CI/CDAzure DevOps
DatabasesExpert
MySQLPostgreSQLRedisSQL ServerOracleMongoDBSnowflake
Messaging & AsyncAdvanced
Apache KafkaRabbitMQActiveMQEvent-Driven PatternsBackground Jobs
Security & TestingAdvanced
OAuth2 / JWTRBACJUnit / MockitoPHPUnitJestTDDSwagger / OpenAPIOWASP Basics

Skill Breadth

AI & Machine LearningAdvanced
BackendExpert
FrontendAdvanced
Cloud & DevOpsAdvanced
DatabasesExpert
Messaging & AsyncAdvanced
Security & TestingAdvanced
Work Experience

Career

SFI Group Ltd

Current

Sr Full Stack Developer

Toronto, ON, Canada

Enterprise Operations Platform

  • Owned end-to-end development of full-stack enterprise applications covering quoting, scheduling, payroll, inventory, and warehouse operations
  • Designed and developed backend services and RESTful APIs using Java and PHP (Laravel), consumed by React and Vue frontends
  • Maintained and extended a React Native mobile application used by field operations staff for real-time job tracking and scheduling
  • Re-architected legacy monolithic modules into modular, maintainable components — significantly improved system stability and scalability
  • Optimized SQL queries, indexes, and schemas — reduced high-latency database operations by ~30%
  • Built asynchronous background processing pipelines for reporting, exports, and long-running operational workflows
  • Implemented Redis caching strategies to improve performance of frequently accessed data
  • Integrated cloud services across AWS (EC2, S3) and Azure; containerized applications using Docker and Kubernetes
  • Prototyped LLM-assisted features for internal tooling — automated summarization of operational reports and natural-language search over job records
JavaPHP (Laravel)PythonReactReact NativeVue.jsMySQLPostgreSQLRedisAWSAzureDockerKubernetesLLM APIsCI/CD

SFI Group Ltd

Current

Sr Full Stack Developer

Toronto, ON, Canada

Enterprise Operations Platform

  • Owned end-to-end development of full-stack enterprise applications covering quoting, scheduling, payroll, inventory, and warehouse operations
  • Designed and developed backend services and RESTful APIs using Java and PHP (Laravel), consumed by React and Vue frontends
  • Maintained and extended a React Native mobile application used by field operations staff for real-time job tracking and scheduling
  • Re-architected legacy monolithic modules into modular, maintainable components — significantly improved system stability and scalability
  • Optimized SQL queries, indexes, and schemas — reduced high-latency database operations by ~30%
  • Built asynchronous background processing pipelines for reporting, exports, and long-running operational workflows
  • Implemented Redis caching strategies to improve performance of frequently accessed data
  • Integrated cloud services across AWS (EC2, S3) and Azure; containerized applications using Docker and Kubernetes
  • Prototyped LLM-assisted features for internal tooling — automated summarization of operational reports and natural-language search over job records
JavaPHP (Laravel)PythonReactReact NativeVue.jsMySQLPostgreSQLRedisAWSAzureDockerKubernetesLLM APIsCI/CD

Enlacecode Pvt Ltd

Full Stack Developer

Ahmedabad, India

Healthcare Software Modernisation Platform

  • Led migration of a legacy healthcare management system to a modern multi-stack architecture, replacing deprecated modules with maintainable, scalable services
  • Designed and developed backend APIs using Java, Node.js, and PHP (Laravel) to power migrated clinical workflows and patient data pipelines
  • Architected modular backend components separating business logic, reporting workflows, and role-based access control across healthcare user roles
  • Built SQL-optimised APIs applying indexing and query refactoring to support large patient record datasets and time-based reporting
  • Developed interactive frontend components using Vue.js and JavaScript for clinical dashboards during the migration phase
  • Implemented RBAC and secure API authorisation for physician, admin, and staff roles; documented APIs using Swagger/OpenAPI
JavaPHP (Laravel)Node.jsVue.jsJavaScriptMySQLREST APIsSwaggerPHPUnitPostmanGitBitbucket CI

Enlacecode Pvt Ltd

Full Stack Developer

Ahmedabad, India

Healthcare Software Modernisation Platform

  • Led migration of a legacy healthcare management system to a modern multi-stack architecture, replacing deprecated modules with maintainable, scalable services
  • Designed and developed backend APIs using Java, Node.js, and PHP (Laravel) to power migrated clinical workflows and patient data pipelines
  • Architected modular backend components separating business logic, reporting workflows, and role-based access control across healthcare user roles
  • Built SQL-optimised APIs applying indexing and query refactoring to support large patient record datasets and time-based reporting
  • Developed interactive frontend components using Vue.js and JavaScript for clinical dashboards during the migration phase
  • Implemented RBAC and secure API authorisation for physician, admin, and staff roles; documented APIs using Swagger/OpenAPI
JavaPHP (Laravel)Node.jsVue.jsJavaScriptMySQLREST APIsSwaggerPHPUnitPostmanGitBitbucket CI

Iworld Pvt Ltd

Software Developer

Gujarat, India

Trading & Analytics Platform

  • Designed and developed full-stack trading and analytics applications using Java and ASP.NET (C#) backend services
  • Built RESTful APIs and backend modules for trade data ingestion, portfolio tracking, and real-time analytics reporting
  • Implemented server-side business logic for position management, P&L calculations, and operational reporting workflows
  • Developed web-based frontend interfaces using JavaScript, HTML, and CSS for traders and operations teams
  • Designed and optimised SQL queries, stored procedures, and schemas for high-volume trade data and time-series analytics
  • Assisted with refactoring legacy components to reduce technical debt and collaborated with senior engineers on architecture decisions
JavaASP.NETC#JavaScriptHTMLCSSMySQLSQL ServerREST APIsMVC Architecture

Iworld Pvt Ltd

Software Developer

Gujarat, India

Trading & Analytics Platform

  • Designed and developed full-stack trading and analytics applications using Java and ASP.NET (C#) backend services
  • Built RESTful APIs and backend modules for trade data ingestion, portfolio tracking, and real-time analytics reporting
  • Implemented server-side business logic for position management, P&L calculations, and operational reporting workflows
  • Developed web-based frontend interfaces using JavaScript, HTML, and CSS for traders and operations teams
  • Designed and optimised SQL queries, stored procedures, and schemas for high-volume trade data and time-series analytics
  • Assisted with refactoring legacy components to reduce technical debt and collaborated with senior engineers on architecture decisions
JavaASP.NETC#JavaScriptHTMLCSSMySQLSQL ServerREST APIsMVC Architecture
Featured Work

Projects

Production systems, mobile apps, and machine learning models — built for real environments, not demos.

Humber College · Competition

Top 5

race finish · Jan 2025

AWS DeepRacer — Autonomous Racing Model

Trained an autonomous driving model with reinforcement learning on AWS SageMaker. Iterated on reward function design and hyperparameters until the car stopped fighting the track — and finished top 5.

PythonReinforcement LearningAWS SageMakerReward ShapingHyperparameter Tuning
AI Lab · In Progress

RAG

LLM + vector search

RAG Document Assistant

Full-stack document Q&A assistant: upload documents, ask questions in plain English, get cited answers. Embeddings + pgvector semantic search with an LLM answering layer and streaming UI.

Next.jsTypeScriptPythonClaude APIPostgreSQLpgvectorEmbeddings
SFI Group Ltd

~30%

DB latency reduction

Enterprise Operations Platform

End-to-end enterprise platform covering quoting, scheduling, payroll, inventory, and warehouse operations for a multi-site operations company — plus LLM-assisted report summarization.

JavaPHP (Laravel)ReactVue.jsRedisPostgreSQLDockerAWSLLM APIs
SFI Group Ltd

Geofenced

mobile clock-ins

Timewise — GPS Time-Tracking App

React Native HR time-tracking app with GPS geofencing — employees can only clock in on-site. Laravel REST API with queue workers processing location events asynchronously, plus a Vue.js admin dashboard.

React NativeLaravelPostgreSQLQueue WorkersVue.jsGPS APIs
Enlacecode Pvt Ltd

Multi-role

RBAC with real-time data

Trading & Analytics Platform

Multi-stack real-time trading analytics platform with dynamic dashboards, RBAC access control, and SQL-heavy time-series reporting pipelines.

JavaNode.jsPHP (Laravel)Vue.jsMySQLSwaggerREST APIsBitbucket CI
Machine Learning

The AI Lab

Where I train models, break them, and train them again. Applied ML — not slideware.

Case Study · AWS DeepRacer

Teaching a car to drive itself — all the way into the top 5.

In January 2025 I entered Humber College's AWS DeepRacer competition: train a 1/18th-scale autonomous car to lap a track using nothing but reinforcement learning. No steering wheel, no if-else driving logic — just a reward function and a lot of training hours on AWS SageMaker.

The interesting part wasn't the racing. It was watching reward shaping change behaviour: too much reward for speed and the car flew off corners; too much for centering and it crawled. I iterated on the reward logic and hyperparameters — learning rate, discount factor, training time — until the model stopped fighting the track.

Final result: a top-5 finish, and the realization that I enjoy debugging model behaviour as much as debugging code. That race is why this section of the site exists.

Top 5Race finish
RLReinforcement learning
SageMakerTraining infrastructure
reward_function.pyPython
1def reward_function(params):2"""Reward fast, centered driving. Punish wobble."""3track_width = params['track_width']4distance_from_center = params['distance_from_center']5speed = params['speed']6steering = abs(params['steering_angle'])8# stay tight on the racing line9marker = 1 - distance_from_center / (track_width * 0.5)10reward = max(marker, 1e-3)12# carry speed through the straights13if speed > 2.5 and steering < 10:14reward *= 1.516# zig-zagging is slow. punish it.17if steering > 20:18reward *= 0.820return float(reward)
simplified from the version that made the top 5

AWS DeepRacer — Top 5 Finish

Humber College · Jan 2025

Trained an autonomous racing model with reinforcement learning on AWS SageMaker. Reward shaping + hyperparameter tuning over dozens of training runs.

Microsoft Certified: Azure Fundamentals (AZ-900)

Microsoft

Cloud concepts, core Azure services, security, and governance — the foundation under every model I train and every workload I ship.

DMG MORI Leadership Scholarship

DMG MORI Group · Jan 2025

Industry-sponsored scholarship recognizing leadership and a strong vision for the future of manufacturing and technology.

On the bench right now
  • RAG document assistant — LLM + pgvector semantic search over real documents
  • LLM-assisted report summarization for operational data
  • Anomaly detection experiments on time-series clock-in data
Education

Academic

Post-Graduate

PG Diploma in Information Technology Solutions

Humber College

Toronto, ON, Canada
Under-Graduate

Bachelor of Computer Applications (BCA)

Gujarat University

Ahmedabad, Gujarat, India
Credentials✦Microsoft Certified: Azure Fundamentals (AZ-900)GitHub Student Developer Pack · DMG MORI Leadership Scholarship
Get In Touch

Let's build
something
serious.

Open to software engineering and AI/ML engineering roles — full stack, backend, or applied machine learning. Based in Toronto — remote-ready.

Ghanshyamsoni682@gmail.com
linkedin.com/in/ghanshyamsoni
github.com/GhanshyamSoni68

Response time: Usually within 24 hours

Protected by reCAPTCHA · Rate limited · Honeypot filtered

0/5000