Forward Deployed Engineer

Building software where the work actually happens.

Based in

Toronto, Canada

Education

Bachelor of Computer Science

Focus

Full-stack products, applied AI, and owning what ships.

Scroll

(01) — Selected Work

Shipped, used, and still running.

A live client platform I build and run on my own, four award-winning hackathon builds, and a live look at what I’ve been shipping on GitHub.

01 Live

Client · 2026 — Present

Central Auto Sales

Production platform for a multi-location used car dealership. Searchable inventory, side-by-side comparison, an admin dashboard and a daily new-arrivals newsletter. I’m the only engineer: I built it, ship its features, and stay on call.

Role: Full-Stack Engineer
Scope: Design, build, run
Stack: Next.js · Supabase · Vercel

Hackathon wins

04 projects

Hackathon winner

Magic Quill

Real-time lecture transcription that turns speech, slides, diagrams and equations into clean, organized notes, exported as Markdown and PDF.

Next.jsGeminiOCRLaTeX

Hack49 · Winner

ALI

An AI companion app that screens for early signs of Alzheimer’s through natural speech, while keeping elderly users company and less isolated.

React NativeFastAPIPyTorchWav2Vec

HackTheValley 9 · Winner

TrashCam

Real-time waste sorting through your camera. Object recognition with Google Cloud Vision and COCO-SSD tells you which bin each item belongs in.

Next.jsTypeScriptPostgreSQLGemini

GenAI Genesis (Google) · Winner

SageAI

An AI diagnostic tool that reads written symptoms with BERT and turns uploaded images into descriptions with Google Imagen, reaching over 90% accuracy across 41 diseases.

ReactNode.jsPyTorchVertex AIMongoDB

GitHub activity — last 12 months

— contributions

@wiledw ↗

(02) — About

Engineering is an act of trust.

I’m William Edwardo Gunawan, a Forward Deployed Engineer in Toronto. I graduated from Toronto Metropolitan University with a Bachelor of Computer Science with Distinction, shipped at startups like Fiscal AI and Postilize, and build and run production software for clients. I work directly with the teams who use what I build.

Off-screen — music, games, travel, food

Philosophy

01

Start with the user, not the stack.

I sit close to the people who will use the software. Their workflow decides the architecture, and the tools come after.
02

Ship, then sharpen.

Small releases, real feedback, steady iteration. A working version in front of people beats a perfect plan nobody has seen.
03

Own it after launch.

Monitoring, on-call, fixing what breaks. The job doesn’t end at deploy. That’s where people start depending on it.

Capabilities

Languages

TypeScript
JavaScript
Python
Java
SQL
C#

Frameworks

Next.js
React
Node.js
FastAPI
Django
TensorFlow

Platforms

AWS
Azure
GCP
Supabase
Docker
Redis
Claude

(03) — Experience

Six years of building in teams.

Embedded with client teams to scope real problems on the ground, then build and ship the software that solves them.

greenlightconsulting.com
Sole engineer on the dealership’s production website: designed, built and deployed it end to end.

Built searchable multi-location inventory with advanced filters and side-by-side vehicle comparison.

Built an admin dashboard for vehicles, locations and image uploads, plus a daily new-arrivals newsletter and push notifications.

On call for production issues and new feature requests.

Next.jsTypeScriptSupabaseVercelResend
central-autosales.com
Raised the precision of AI-detected sales signals, finding and removing a duplicate class about 35x larger than any fixed before and reaching 95.7% precision on litigation filtering, by building URL-normalized and entity-overlap dedup gates plus GPT-4.1 verifiers validated through offline replay and blind grading.

Turned SEC filings that ran for hours and re-queued night after night into extractions of 1.4 s or less on the largest filings, by rebuilding EDGAR ingestion on BullMQ with fan-out queues, per-filing retries and a shared rate limiter, and replacing a backtracking regex with a linear scan proven equivalent by 20,000-case fuzz tests.

Cut timing-out feed searches for the largest customers from 43 s to about 14 s by routing queries to indexed lookups, and built an Atlas Search path that reads 26–265x fewer rows with no loss of precision.

Built CRM data cleansing across Salesforce, HubSpot and Foundation 365 that suggests and applies contact fixes (job changes, name corrections, archiving and enrichment), holding back about 15% of employer-change suggestions that contradicted email evidence and adding readable causes for the 59% of failed CRM writes that previously had none, by bringing the connectors to parity, consolidating and deduplicating suggestions, flagging previously rejected fixes for a six-month recheck, and leading an apply-failure diagnostics program.

Launched a new public-agency signal product end to end, now producing signals from 19 utility-district and agency sources, by building Firecrawl crawling, local PDF and OCR parsing, GPT extraction, and self-serve source onboarding that learns a site’s document pattern from a single example link.

Built a hiring-signal pipeline that tells law-firm partners when companies they follow are recruiting for 14 senior legal, finance and executive roles across 36 locations in the US, Canada and Europe, by ingesting public job postings on admin-configured regional schedules that update live over Redis pub/sub, filtering false positives with title heuristics plus an LLM check reserved for ambiguous titles, and caching each decision in Redis so repeat titles aren’t re-classified on every thrice-daily run.

Node.jsMongoDBBullMQRedisOpenAIAtlas SearchSalesforceHubSpot
postilize.com
Built a scalable Python ETL pipeline that extracts and processes SEC EDGAR filings for 4,000+ U.S. companies, covering up to 40 years of history.

Reached 98% extraction accuracy for employee counts and stock splits with a multi-step, chained Gemini LLM pipeline.

Improved parser performance by 30% through weekly batch testing with analysts, 100 companies at a time.

Automated data refreshes with AWS Lambda, S3 and Parquet, so new filings update the datasets with no manual work.

PythonGeminiAWS LambdaS3Parquet
fiscal.ai
Worked with project management and marketing to turn Figma designs into scalable web applications.

Set front-end testing standards with Jest and React Testing Library, reaching over 90% coverage on critical components and reducing post-deploy bugs.

ReactJestReact Testing LibraryFigma
Helped design and build a scalable full-stack platform with C#, ASP.NET, TypeScript, React and AWS RDS.

Designed a RESTful API between front end and back end and ran weekly testing with beta users, improving performance by 30% and halving average response time.

Wrote unit tests for controllers and services with Moq and xUnit, improving code coverage by 70%.

C#ASP.NETReactAWS RDSxUnit
Built a scalable full-stack application with a React front end and a Node.js back end.

Worked with cross-functional teams to add testing to the CI/CD pipeline, cutting deployment time by 50% and production issues by 20%.

Set up and maintained the MongoDB database for storing and querying application data.

ReactNode.jsMongoDBCI/CD
cove.dev
Built and maintained RESTful APIs with Flask and front-end components with React.

Integrated AWS S3 for file storage, using pre-signed URLs to keep uploads secure.

Set up a GitHub Actions CI/CD pipeline for automated testing and linting, reducing deployment friction by 40%.

FlaskReactAWS S3GitHub Actions
rapiertechnology.co.id