Hey there, my name is

Evan Lamb

Backend Software Engineer

Software Engineer

January 2026 – April 2026

Architected an automated data pipeline using Python and MySQL to extract and synchronize Salesforce records into New Relic, delivering a centralized dashboard that enhanced system observability for the Integrations team.
Engineered a Co-pilot Studio agent utilizing natural language processing to automate the generation of standardized Jira tickets, eliminating manual administrative overhead and accelerating issue resolution.
Optimized business intelligence workflows by developing SQL queries within Salesforce Marketing Cloud, extracting critical performance metrics to drive data-informed decisions for senior business stakeholders.
Orchestrated meetings translating business challenges from the sales team into actionable software requirements.
Secured and deployed a custom Python app pipeline to the cloud using Kubernetes and Jenkins CI/CD workflows, enforcing security standards by performing Snyk scans and GitGuardian secret alerts prior to deployment.

PythonMySQLSalesforceNew RelicCo-pilot StudioSQLKubernetesJenkinsSnyk

Software Developer

May 2025 – August 2025

Developed a forecasting agent using Python that ingested and processed time-series data from 91 Kubernetes hosted services, enabling an estimated $400K in annual savings for a single team and demonstrating POC.
Accelerated cross-team adoption by developing a web UI and backend on-boarding workflow that streamlined integration of the forecasting agent, reducing setup time for new teams using Flask REST APIs.
Reduced MAPE (from 10% to 7%) by using LangGraph and LangSmith for performance monitoring, data fetching, cyclic reasoning, and data cleansing, enabling more reliable forecasts and better-informed decisions.
Achieved 5% MAPE with only 100 data points by creating an "Iterative Calibration" method simulating LLM backpropagation with mini chain-of-thought runs, reducing forecast bias and improving prediction accuracy.
Secured an initial and follow-up meeting with EVP of Core Banking & Payments Technology by presenting potential enterprise-wide benefits, leading to deeper evaluation of performance and cost savings.
Documented data workflows and pipeline setup in Confluence, creating reusable references that reduced onboarding time across 3 teams and reinforced best practices in data pipeline management.
Improved service stability by 20% across 91 microservices by optimizing CI/CD pipelines (GitHub Actions, SonarQube, Docker), strengthening deployment resilience and aligning with cloud-first DevOps practices.

PythonFlaskKubernetesLangGraphLangSmithGitHub ActionsSonarQubeDockerConfluence

An AlphaZero-style game-playing agent that learns purely from self-play, no human game data.

PyTorchPythonNumPyFastAPI

Implemented an AlphaZero-style game-playing agent from scratch that learns purely through self-play with no human game data, integrating Monte Carlo Tree Search with a policy/value neural network.
Engineered the full self-play training loop, game generation, a symmetry-augmented replay buffer, batched network updates, and versioned, resumable checkpointing with a NaN/inf loss guard to prevent poisoned checkpoints.
Developed an evaluation harness (arena match runner, generation-vs-generation win rates, anchored Elo with CSV/JSON export) that quantitatively confirmed each agent beats its baselines.

Detect and filter non-educational videos from your YouTube recommendations, instantly, with classical machine learning.

PythonJavaScriptHTMLCSSChrome Extension API

Classifies YouTube videos as educational or non-educational from their metadata to filter your YouTube feed in real time.
Executes inference entirely client-side in the browser using a lightweight logistic regression and TF-IDF model.
Built a Python data pipeline that handles data collection, labeling, training, and exporting the model to a browser-ready format.

Citation-grounded RAG system for Canadian refugee law

Next.jsTypeScriptFastAPISnowflakeGeminiLangChain

Engineered a full-stack RAG system for Canadian refugee law, delivering citation-grounded legal responses (ConHacks — Snowflake Track Winner).
Built ingestion pipelines curating 3,000+ tribunal decisions and 70+ statutes in Snowflake; implemented chunking, metadata enrichment, and top-k semantic retrieval using LangChain and Snowflake Cortex Search.
Architected a routed RAG pipeline with query rewriting and Gemini-based intent classification, dynamically selecting retrieval sources; enforced citation-only outputs via grounded prompting to reduce hallucination.

A browser-based, AI-native Blender alternative that democratizes 3D game development with integrated generation, modeling, and animation.

TypeScriptReact 19Three.jsViteZustandGemini APIMeshy API

Engineered an accessible 3D scene editor with a hierarchical structure and customizable dockable windows to collapse the game development pipeline into a single workspace.
Integrated Google Gemini and the Meshy API to power primitive model blockouts, image-to-3D generation, and natural-language animation selection.
Developed a custom edit-mode using Three.js for vertex/edge/face manipulation and GLB export, earning Best Use of VR and the Gemini API category at HackCanada.

Portfolio and blog built with Next.js

Next.jsReactTypeScriptTailwind CSSVercel

Designed and developed a personal portfolio site to showcase work experience, projects, and blog posts.
Built with Next.js App Router and Tailwind CSS for a fast, responsive, and accessible experience.
Deployed on Vercel with automatic preview deployments for every pull request.