Why Maya exists
Technical skills open doors. The confidence to explain those skills helps a job seeker walk through them. For visually impaired learners, interview preparation must be accessible, encouraging and grounded in the realities of how they work. That is why we built Maya.
Since 2004, Vision-Aid has been enabling, educating and empowering visually impaired people in India to live with independence and dignity. Vision-Aid serves over 100,000 beneficiaries annually. In the United States, Vision-Aid is run by an all-volunteer team. In India, its programs operate across more than 85 locations in collaboration with leading eye hospitals, blind schools and community organizations.
Helping learners build specialized skills and find meaningful work is central to this mission. Ability Bazaar is Vision-Aid’s platform connecting visually impaired learners and job seekers with new digital opportunities, particularly in artificial intelligence. Launched in July 2026, it brings together emerging technology and pathways to meaningful work. To strengthen this pathway, Vision-Aid is launching the Ability Fellowship Program, combining accessible technical training, practical projects, mentoring, internship experience and placement support across Digital Accessibility Testing, AI Full-Stack Development and Digital Marketing.
Maya addresses a crucial step between learning and employment: preparing for the interview. It gives visually impaired job seekers a supportive place to practise explaining what they know, how they solve problems and the value they bring to an employer. Voice-first interaction, keyboard and text alternatives, practical questions, encouraging feedback and retries help learners strengthen their answers at their own pace.
Built around the needs of visually impaired learners, Maya complements trainers and mentors by making independent practice available between sessions. Ability Fellows are expected to use Maya to prepare and hone their interview skills when applying for jobs. The purpose is simple: help talent find its voice, build interview confidence and take the next step toward meaningful work.
What Maya does
Maya is a free, open-source interview coach designed to help visually impaired job seekers build confidence and communicate their skills. Available through Ability Bazaar, it is being developed as a lasting resource for learners, trainers and placement teams—not simply a competition demonstration.
Learners choose a practice area, hear a realistic interview question, and answer by voice or typing. Maya provides encouraging, practical feedback and an opportunity to retry, helping them strengthen their answers at their own pace. Learners can ask for repetition, hints or a sample-answer structure, then reflect on their progress through a session summary.
Maya supports Digital Accessibility Testing, Full-Stack Development and Digital Marketing. Its questions connect technical knowledge with practical workplace situations, helping candidates explain how they approach tasks, solve problems and contribute to an employer. An optional “Surprise me” round uses Amazon Bedrock to generate fresh questions and feedback, adding variety alongside the regular practice.
The Build, Ship, Shape: Amazon Developer Hackathon inspired and accelerated Maya’s development, including its Alexa+-style conversational experience. Its purpose extends beyond the event: to make accessible interview preparation an ongoing part of Vision-Aid’s training-to-employment pathway. With continued input from learners and trainers, Maya is designed to deliver lasting value to visually impaired people preparing for meaningful work.
What makes it specialized
Maya builds on Vision-Aid’s years of experience training visually impaired learners and helping them prepare for employment. It takes that experience a step further through AI, voice technology and accessible design, turning practical know-how into freely available interview practice.
The specialization shapes both the experience and the content. Learners can practise independently by voice, keyboard or typing. Questions help them explain how they apply their skills in real workplace situations. For example, a candidate preparing for Digital Accessibility Testing can rehearse describing how they use a screen reader and keyboard to test a registration form, report an accessibility barrier and verify a correction.
Maya’s questions and coaching guidance can grow with contributions from trainers and placement teams. AI-generated surprise questions add fresh challenges alongside structured practice, helping learners apply their knowledge in new situations. Encouraging feedback, hints and retries give them room to improve at their own pace.
This combination of practical training experience and emerging technology gives Maya its distinctive focus: helping visually impaired job seekers explain their capabilities, build confidence and prepare for meaningful work.
Improved through trainer and team feedback
The team learned many valuable things while evaluating Maya.
- Learners need a comfortable place to practise. Our placement officer observed that students who speak confidently in groups can become nervous in one-to-one interviews. Her feedback reinforced the value of independent practice, with time to think and try again.
- Feedback needs to be specific and encouraging. We improved Maya’s responses to explain what the learner covered and suggest a useful next step. Learners can request more detail, ask for a hint, or retry.
- Mentioning a technical term does not necessarily demonstrate understanding. Team members deliberately tried incomplete and misleading answers. We tightened the practice checks and removed percentage scores that could give a misleading impression of accuracy.
- Learners need control over the conversation. We added clearer options to repeat, skip, get help and improve an earlier answer. Typed answers remain available for editing during a retry.
- The opening needs to be clear and responsive. Maya now introduces all three practice areas, identifies the selected one and explains how to switch. We also prepared the welcome audio in advance to reduce startup delay.
- Trainers bring essential workplace knowledge. Their feedback helped us strengthen practical questions, including how candidates use assistive tools at work, and expand the Digital Marketing practice set.
Why Alexa Plus
Maya is designed for visually impaired job seekers who often find voice and audio the most natural way to use technology. That makes Alexa a natural fit: learners can hear a question, speak their answer, and get coaching through a familiar conversational experience.
The current web app demonstrates this Alexa+-style interaction and provides working interview practice today. We plan to develop and submit an Alexa skill for certification so Maya can reach Alexa-enabled devices and make accessible interview coaching more widely available.
This builds toward our longer-term goal: helping visually impaired job seekers practise independently, build confidence, and prepare for meaningful work.
How we built it
We built Maya around a simple conversation: hear a question, speak an answer, receive coaching and try again. This is the experience we intend to bring to Alexa-enabled devices.
The working web app demonstrates that conversation and also offers an accessible on-screen interface. Learners can read questions and feedback, navigate by keyboard or type answers when preferred. These options broaden access while keeping Maya voice-first.
We built the web interface with HTML, CSS and JavaScript, and the backend with Python and Flask. An MCP server manages the interview session, including questions, feedback, retries and progress. Questions and coaching guidance live in an editable bank so they can grow with contributions from trainers.
Maya speaks in a consistent female Indian-English voice. Audio controls prevent overlapping playback and keep longer feedback playing in sequence.
Regular practice uses structured questions and rubric-based coaching. The optional “Surprise me” round uses Amazon Bedrock with Meta’s Llama model to generate fresh questions and feedback. Learners can return to their regular practice at any time.
AWS credentials stay on the server. Timeouts and usage controls help manage requests. If the AI service is unavailable, regular practice remains available.
We tested the interview flow, audio controls and optional AI round through automated tests and live checks. Feedback from trainers and team members helped us refine the experience.
Maya’s standalone code, setup instructions and tests are available in a public, open-source repository so others can study, adapt and build on the app.
What is new for this hackathon
Ability Bazaar predates this event. Maya is the new, separately open-sourced interview-practice app developed during the hackathon period: the three-track practice flow, DAT/ARIA content, stateful coaching tools, feedback-and-retry loop, speech interface, cloud-voice integration, tests and standalone packaging. We are not resubmitting the earlier marketplace as a new product.
Challenges we faced
Building a voice-first coach taught us how much the transitions matter. Learners need to know when Maya is speaking, when she is listening and what happens next. We worked on clearer prompts and handoffs to make the conversation easier to follow.
Testing uncovered overlapping audio and inconsistent voices, so we added controls to keep playback clean and consistent.
Microphone behavior varied across browsers and phones. We corrected a setting that blocked microphone access and added clearer error messages. Long spoken answers still need work because recognition can stop before someone finishes. Our focus is on giving enough time and making each handoff clear.
Another challenge was making feedback useful. We refined the coaching to acknowledge meaningful short answers, recognize details already provided and suggest practical ways to improve.
When adding generative AI, we wanted to preserve the regular practice experience that testers liked. We introduced the optional Bedrock-powered ‘Surprise me’ round, giving learners fresh challenges while keeping their familiar practice flow available.
These lessons continue to guide development: clear spoken instructions, useful feedback and learner control at every step.
Accomplishments and potential impact
Maya is a working app, freely available through Ability Bazaar. Visually impaired job seekers can practise interview answers, receive supportive coaching and try again at their own pace.
We brought together accessible design, practical training knowledge and an optional AI-powered practice round. We also published Maya’s code so trainers and other organizations can adapt it for their communities.
Vision-Aid’s training network and the Ability Fellowship Program provide a clear path for putting Maya to use. Fellows are expected to use it between training and mentoring sessions to strengthen how they explain their skills, describe their work and prepare for interviews.
As we introduce Maya to learners, we will assess its usefulness through trainer feedback, the quality of practice answers and improvement across retries. Our goal is to make confident, independent interview preparation available to more visually impaired job seekers.
What’s next for Maya
Our immediate priority is making longer spoken answers more reliable, so learners have time to finish before Maya responds.
We will work with fellows and trainers to expand the question banks, review coaching quality and improve response times. Regular practice uses structured rubrics; the optional AI round adds fresh questions and feedback that we will continue evaluating with trainers.
We will also evaluate evolving AI models available through Amazon Bedrock, comparing the usefulness of their coaching, response speed and cost to find the best fit for our learners.
We also plan to give staff tools to review usage and contribute questions and coaching guidance, helping Maya grow with Vision-Aid’s training experience.
Alongside these improvements, we plan to develop and test the Alexa skill and pursue certification, bringing Maya’s voice-first coaching to Alexa-enabled devices.
Try it
Visit Ability Bazaar’s Apps page and choose Maya, or open Maya directly. Select a practice area, start voice practice and allow microphone access.
Answer a question, listen to Maya’s coaching, then try again with a stronger example. For a fresh challenge, say “Surprise me” to experience questions and feedback powered by Amazon Bedrock. Say “Back to practice” to return to your regular interview session.
Maya is free to use, and its source code is publicly available on GitHub under the MIT license. The demonstration video uses a simulated learner voice to show the working app. Please avoid sharing confidential or personal information in practice answers.
Log in or sign up for Devpost to join the conversation.