Skip to content
View shi1720's full-sized avatar

Block or report shi1720

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
shi1720/README.md

Shivam Gupta. Applied AI, thoughtful products, working software.

Shivam Gupta

Applied AI engineer · Product builder · Founder, Siloed

I turn “someone should build this” into “try this link.”

I build AI products from the first customer conversation through architecture, implementation, evaluation and deployment. My work spans education, enterprise software and small experiments with a point of view. I care about useful products, clear interfaces and evidence that the system works.

LinkedIn · Email · Dubai / Delhi

Selected work

A coding benchmark needs trustworthy tests before it can judge an agent. Python orchestration, five language environments, broken and golden controls, reproducible reports.

A job-search CRM for applications, follow-ups and the work between them. FastAPI, React and Google Cloud, with grounded AI drafts, retry-safe imports and scheduled reminders.

Accessibility repairs with evidence that they helped. Coding agents, independent scans, content-integrity checks and a human review path.

A personality quiz that shows its working. TypeScript scoring, a Python reference implementation, private local answers and browser-generated share cards.

Travel recommendations leave out how a city actually felt. Anonymous trip reports, a 3D globe, weekly reveals and a persistent database. Live app; source stays private.

Browse the public project index for smaller tools, experiments and their status.

Beyond the public repos

  • Khoros: led AI and analytics engineering for the IRIS AI modernization, working across a team of five engineers and a designer.
  • 2 Hour Learning: helped build learning applications serving 5,000+ learners and co-developed Incept, a content-generation pipeline that saved $1M+ annually.
  • Siloed: founded an AI product consultancy and delivered work for 10+ clients across finance, healthcare, SaaS and other industries.

These are contributions to professional teams and client projects. Their implementation remains private.

How I work

Start with the user's problem. Make a small version work. Measure failure cases. Improve the architecture where the evidence calls for it. Document the decisions someone else will have to live with.

My usual tools are Python, TypeScript, FastAPI, React, PostgreSQL and Google Cloud, with LLM APIs and evaluation harnesses where they earn their place. I use AI-assisted development and take responsibility for the result.

Outside work, I travel, follow interesting detours and occasionally turn them into another project.

Popular repositories Loading

  1. speechace-proxy speechace-proxy Public

    Keep pronunciation-scoring credentials server-side while forwarding bounded audio uploads to Speechace.

    JavaScript

  2. sat-pdf-to-csv sat-pdf-to-csv Public

    Extract text from SAT question PDFs and use a model-assisted parsing step to produce a reviewable CSV question bank.

    Python 1

  3. ap-frq-review ap-frq-review Public

    Review AP free-response questions in batches using configurable evaluation prompts, parallel requests and downloadable feedback.

    Python

  4. ap-article-review ap-article-review Public

    Review batches of AP learning articles against configurable quality prompts, inspect model feedback and export review results.

    Python

  5. math-question-editor math-question-editor Public

    Edit, preview and export math question banks without manually restructuring each question. Supports JSON, tabular review and PowerPath-compatible output.

    Python

  6. ws-eng-conduit-ai-assessment ws-eng-conduit-ai-assessment Public

    Forked from trilogy-group/ws-eng-conduit-ai-assessment

    An upstream assessment fork retained for provenance. The default branch contains assessment instructions, not an original standalone application.