Inspiration
You can't aim for a job you've never seen. Many students, especially first-generation students, have never watched anyone do the jobs they're applying for, and they walk into interviews nobody coached them for. Deaf and hard-of-hearing students face both gaps plus an accessibility one: most career content is audio-first.
What it does
Day One turns any job into a short movie trailer, then lets you practice getting it.
- Type a job, or paste a real job posting. Gemini reads the posting and researches the company.
- You direct: pick a genre and dress Bolt, our robot mascot, for the part.
- While Bolt runs to work in a mini game, the trailer is made: six scenes of a day in that job, narrated by Bolt with word-by-word captions, and signed in American Sign Language.
- The credits show the real path into the job and what it costs to get there.
- The Screen Test: Bolt reads your resume and interviews you out loud and in sign. Answer by voice or by typing. A vague answer gets a gentle follow-up, like a real interviewer.
- You get a score on four skills (structure, specifics, relevance, ownership), the fix for each answer, and your progress across attempts at the same job.
- Come back later and Bolt remembers you and suggests what to practice next.
Every screen works in English, Spanish, Hindi, Arabic (right to left) and Vietnamese.
For the Future of Work: Day One makes the path to work more accessible. Anyone can see a real day in a job, learn the path and cost to get there, and practice the interview, in their own language or in sign language, for free.
Bolt, our mascot
Bolt hosts every step: he greets you, wears the outfit you choose, runs to work while the trailer renders, acts out the job, lip-syncs on his visor, signs each line, becomes your interviewer, and welcomes you back. A friendly robot instead of a stern panel takes the pressure off a first mock interview.
How we built it
- AWS Step Functions runs the studio: write the script, then a Map state makes the scenes in parallel, then finalize. Lambda + API Gateway serve the API.
- Amazon Polly gives Bolt his voice, with word timings that drive the captions. Amazon Transcribe streams spoken interview answers from the browser, using Cognito guest access.
- Amazon Bedrock (Nova Lite as a text backup), Translate, Rekognition, DynamoDB for job status, S3 + CloudFront for hosting.
- Gemini API researches the posting and company, writes each scene, reads the resume, asks and scores interview questions against a fixed rubric, and turns lines into ASL gloss.
- MongoDB Atlas stores practice sessions per job (the job description, every attempt and per-skill scores), caches finished trailers, and powers the "Now showing" board.
- Supermemory remembers what each visitor watched and practiced, for Bolt's welcome-back line.
- The site is live on a .us domain from GoDaddy Registry: dayonetrailers.us.
- Frontend: React 19, Vite, GSAP. The mini game's music is generated in the browser with Web Audio.
Accessibility and privacy
- Captions on every spoken line. Bolt signs using motion captured from a native ASL signer (the ASLLVD dataset, Boston University). A back-translation check hides any line whose meaning doesn't survive, so those stay captions-only. The signs have not yet been reviewed by a Deaf signer.
- The camera preview never leaves the browser. The resume is used to write questions and is never stored. "Forget me" deletes a visitor's saved practice.
Challenges we ran into
- Turning English into ASL that keeps its meaning, and keeping the signing in sync with the voice.
- Fitting sign generation inside API time limits: we sign all scenes in one batch when the trailer finishes, and cache the results.
- Making interview scores fair and consistent from one attempt to the next, and keeping each job's history separate.
Accomplishments that we're proud of
- A complete loop, live on the web: see the job, learn the path, practice getting it.
- Every line captioned and signed, in five languages.
What we learned
Accessibility has to be designed in from the start. Captions, signing and right-to-left layout touched every screen.
What's next for Day One
- A review by Deaf signers, and more signs.
- More real pay data, and practice for group and technical interviews.
Built With
- amazon-bedrock
- amazon-cognito
- amazon-polly
- amazon-rekognition
- amazon-transcribe
- amazon-translate
- amazon-web-services
- api-gateway
- aws-lambda
- aws-step-functions
- cloudfront
- dynamodb
- gemini
- godaddy
- gsap
- javascript
- mongodb-atlas
- node.js
- react
- supermemory
- vite
- web-audio