I’m Arpit Kumar Jaiswal, a developer from MIT Manipal. I build with Python and TypeScript, working across backend systems, developer tools, and applied AI. My projects focus on making complicated systems easier to understand, use, and verify.
Read the engineering walkthroughs
Make a codebase easier to navigate. A repository analysis workspace connecting import graphs, AST-based metrics, code findings, npm advisories, and saved review snapshots.
Engineering to inspect: relative-import resolution, signed webhook delivery deduplication, and a rules-based review path with an optional AI adapter.
Source · Design decisions · Build & verification
Show the evidence behind a ranking. A React/FastAPI candidate workspace with explainable lexical and skill scoring, excerpts, skill gaps, and a private Resume Lab.
Engineering to inspect: bounded extraction workers, independent resume/JD scoring, transactional uploads, and stale-write protection.
Source · Design decisions · Recorded verification
Product preview — recruiter workspace with synthetic demonstration data
Original screenshot from the project documentation. Example candidates and scores describe the demonstration workspace.
Make data boundaries explicit. A Python SDK and HTTP API for policy-gated encrypted records, tenant-scoped exact search, and authenticated audit chains.
Engineering to inspect: manifest-bound interpretation, atomic record/search-term updates, optimistic revisions, and plaintext release after read-audit commit. Research prototype; not independently security-audited.
Source · Design decisions · Security boundaries · Recorded verification
More work: voice interfaces, computer vision, document search, and developer tools
| Project | What to explore | Stack |
|---|---|---|
| LEXA | macOS-first AI desktop companion with voice recognition and allowlisted local skills | Python · Whisper · PySide6 |
| LeafScan | Leaf-photo classification, account flows, and private scan history | React · Express · MongoDB · PyTorch |
| DevOS | Read-only repository investigations with durable run state and human review | FastAPI · SQLite · MCP |
| DocumentAnalyser | PDF/DOCX extraction, document retrieval, and optional provider-backed answers | FastAPI · scikit-learn · SQLite |
| URLShortener | Link management, aliases, expiry, QR downloads, and private analytics | FastAPI · SQLAlchemy · PostgreSQL |
| veildb | Searchable-encryption research for PostgreSQL applications | Python · cryptography · SQLAlchemy |
| Concord | A workspace for organizing ideas and getting work done | React · TypeScript · Vite |
| CommitSync | Chrome extension for syncing accepted LeetCode solutions to GitHub | JavaScript · Browser APIs |
I care about the behavior behind the interface: clear contracts, consistent state changes, inspectable results, and useful failure states. These three projects apply those ideas in different ways.
The engineering walkthroughs connect each implementation to its design choices, tradeoffs, source revisions, and recorded verification.
Python and TypeScript are my main tools. The rest of the stack comes from the applications linked here.
Select an icon to explore the technology. Project examples are listed below.
Languages & foundations
Web & APIs
Databases & storage
Machine learning & data
Tools & delivery
AI integrations
Stack details and where each tool appears
| Area | Tools | Example repositories |
|---|---|---|
| Frontend | React · TypeScript · Vite · HTML · CSS | NEXORA, TalentOS, Concord |
| APIs & validation | FastAPI · Pydantic · Node.js · Express · HTTPX · MCP | DevOS, NEXORA, LeafScan |
| Persistence | PostgreSQL · SQLAlchemy · Alembic · SQLite · MongoDB · optional Redis | AegisVault, TalentOS, LeafScan, URLShortener |
| ML & data | PyTorch · scikit-learn · NumPy · Pandas · SciPy · Matplotlib · Seaborn | LeafScan, candidate ranking, data-analysis projects |
| Voice & providers | Whisper · PySide6 · OpenAI · Anthropic · optional Ollama | LEXA |
| Delivery & verification | Docker · GitHub Actions · pytest · Playwright · Hypothesis · Ruff · uv | NEXORA, AegisVault, TalentOS |
| Programming foundations | Java · C · C++ · JavaScript · SQL · Verilog | Coursework, algorithms, and programming foundations |
These tools reflect project manifests, documentation, and my stated foundations. Some integrations are optional; each repository describes its implementation scope. Forks are excluded from the language graph below.
A connected map of my public non-fork repositories: project areas on the outside, documented tools in the middle, and primary languages at the core.
Open the full-size graph · Download the interactive explorer · Graph data
Save the explorer HTML file and open it in a browser to search, filter, pan, zoom, and inspect each connection. It works offline; GitHub displays the animated image here.
Data sources and graph methodology
- Calendar: contributions displayed on my public GitHub calendar. The moving light is decorative.
- Monthly chart: contributions grouped into the latest 12 calendar months; the current month is partial.
- Languages: one GitHub-detected primary language per public non-fork repository. This profile and repositories without a detected language are excluded. Counts describe repositories, not lines of code or proficiency.
- Work atlas: includes every public non-fork project repository, including those without a detected language. Project areas are curated; tool links come from reviewed manifests and READMEs. Links describe primary languages or documented tool use, not dependencies between repositories. Motion is decorative.
- Updates: the daily workflow refreshes the checked-in graphics. During GitHub HTTP outages, it warns and preserves the dated snapshot. Snapshot dates remain unchanged until fresh data is fetched successfully.
Open to software engineering roles, internships, and collaborations, especially around developer tools, backend systems, and useful AI applications.