Enter the studio ↗ · Explore the work · Run the agent lab · LinkedIn · Email
I'm Shivam, an applied AI engineer, product builder, and founder of Siloed. I work from the first customer conversation through architecture, implementation, evaluation, and deployment.
A recurring question in my work: what happens when the happy path ends? An agent retries a payment. A clinic can't match a recalled lot. A browser loses the response after saving. Those edges are often where the useful engineering begins.
The studio lets you explore a folded 3D map of my projects, ask a source-grounded AI guide about the work, and run a local agent-reliability experiment. The Python export reproduces what the browser shows.
An earlier creative-coding experiment, the Working Atlas, remains available: an interactive city generated from the same public project collection. How it works →
| If you're interested in… | Open this | What to look for |
|---|---|---|
| Trustworthy agent evaluation | RepoGym · RepoGauntlet | Executable repository tasks, broken and golden controls, reproducible evaluation. |
| Finding the smallest failure | Casecrop | Reduce an event trace while preserving dependencies and the same failure. |
| Agents under pressure | Toolstorm | Deterministic tool failures, side-effect contracts, strict offline replay. |
| Useful voice workflows | Benchback · OffHire | Parts-core returns and equipment closeouts, with transcript-grounded evidence. |
| Software with a human point of view | Unpause · Plot Twist | Resume an unfinished creative project, or meet a personality quiz that shows its working. |
| Repair you can inspect | StepFree | Accessibility fixes with independent scans, content-integrity checks, and review. |
Also worth a look: Mixed Signals maps anonymous experiences onto a globe. OfferLoop makes the work between job applications visible. KindHandoff Guard packages explicit handoff acceptance into a small TypeScript library.
Khoros: Led AI and analytics engineering for the IRIS AI modernization, working with five engineers and a designer.
2 Hour Learning: Helped build 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 implementations remain private.
I’m open to applied AI and product engineering roles, remote work and relocation. Through Siloed, I help teams design and ship AI products, operational agents, internal tools, evaluation systems, and AI adoption programs.
I work across product and engineering: talking to customers, finding the problem worth solving, shaping the roadmap, and turning priorities into work a team can deliver. I stay involved through implementation, launch, and adoption, using feedback to improve the product and connect technical decisions to customer value and commercial goals.
I'm proficient with Cursor, Codex, Claude Code, and Cowork. I use AI tools to research, plan, build, debug, and review work. The skill is in framing the task, supplying the right context, delegating clearly, and checking the result. I take responsibility for the architecture, quality, and product decisions.
My engineering foundation includes Python and TypeScript, React and Next.js, FastAPI, PostgreSQL, Google Cloud and Firebase, alongside LLM integrations and evaluation harnesses. I choose the stack around the product, test how it behaves when things go wrong, and document what the next person needs to know.
Dubai / Delhi · Interested in difficult, useful problems. Let's talk.
Repository READMEs describe each project's scope, setup, and limitations.