I'm Suvendu Kumar Sahoo — an engineer who builds AI systems, full-stack products, and automation that survive contact with real users.
One-line pitch: I turn ambitious experiments into reliable software — with validation, tests, clear architecture, and honest evaluation underneath the demo.
| Catch edge cases before they become incidents. Validate inputs, test failure paths, question surprising results. | Connect APIs, interfaces, models, and data pipelines into workflows people actually use. |
| Move across the stack: product interface → service boundary → model output → user experience. | Software that solves a real problem, explains its behavior, and stays maintainable after the demo. |
identity: Suvendu Kumar Sahoo
mode: AI Builder / Full-Stack Engineer
location: India
focus:
- Agentic AI and LLM applications (structured output, evals, fallbacks)
- TypeScript, React, Next.js, Node.js
- Python, ML, computer vision
- Data workflows, testing, automation
principles:
- Validate at the boundary
- Fail safely and visibly
- Measure before claiming
open_to: [open-source, hackathons, AI experiments, practical product ideas]Each mission links to its source — setup, tests, and status live in the repo.
|
EARTH-001 / THE OUTDOOR DIMENSION Problem: Too much screen time, too little real-world discovery. Approach: Local inference with Gemma 3 4B via Ollama analyzes field images and turns observations into real-world challenges. Engineering focus: validated payloads, safe fallbacks, local-first AI.
|
EARTH-002 / THE CONTRACT DIMENSION Problem: Risky clauses hidden in dense legal documents. Approach: Interpretable clause classification + hybrid retrieval + a LangGraph-orchestrated LLM workflow that structures review findings. Engineering focus: explainability, retrieval quality, orchestrated agents.
|
|
EARTH-003 / THE WORKFORCE WATCHTOWER Problem: Fragmented HR spreadsheets and disconnected workforce signals. Approach: Employee-exit intelligence platform that turns HR data into structured workforce insights across facilities. Engineering focus: auditability, multi-facility workflows, analytics.
|
EARTH-004 / THE DEVELOPER DIMENSION Problem: Repeated dev tasks and fragmented tooling. Approach: One CLI with consistent subcommands, shared API clients, and reusable utilities. Engineering focus: typed validation, ergonomics, composability.
|
| Layer | Stack |
|---|---|
| Web | |
| AI / ML / Data | |
| Tooling |
| Villain | Engineering counter-move |
|---|---|
| Green Goblin — unpredictable AI output | Structured-output validation, deterministic fallbacks, evaluation tests |
| Doctor Octopus — tangled system boundaries | Service contracts, explicit state, modular orchestration, traceable failures |
| Mysterio — misleading metrics | Source-backed measurements, reproducible analysis, stated assumptions |
| Sandman — messy operational data | Validation, reconciliation checks, repeatable pipelines |
| Venom — regressions that return | Regression tests, build gates, careful change review |
| Electro — invisible runtime failures | Useful logs, error boundaries, observable behavior |
The animated masthead is generated from source code in the repository rather than relying on a third-party banner service. It includes:
- Pendulum-style web swinging with six connected movement arcs and web-anchor changes.
- Three independently scrolling skyline layers to create parallax depth.
- Animated window lights, rooftop beacons, clouds, stars, and a moving scan line.
- Comic-book impact bursts at web releases and a compact night-patrol HUD.
- A reduced-motion fallback for visitors who prefer less movement.
The generator creates spiderverse-arena.svg as part of the GitHub Actions workflow. The contribution arena below is a separate visualization driven by your real GitHub contribution grid.
View the animation generator · Run or inspect the animation workflow
The grid is my real GitHub contribution history; the scene around it is a generated comic-book city with a swinging wall-crawler.
LIVE CONTRIBUTION CALENDAR | REFRESH THE ARENA
How the arena is generated
- GitHub Actions generates light/dark contribution heatmaps from real profile activity.
- A repo-owned SVG script adds the web-slinger, strand, skyline, windows, halftone details, and mission HUD.
- Output SVGs publish to the
outputbranch on pushes tomainand a daily schedule.
Manual refresh: open the workflow above → Run workflow on main.
SPOT THE SIGNAL
|
v
TRACE THE ROOT CAUSE --> BUILD A FIX
| |
v v
TEST EDGE CASES <------ VALIDATE THE RESULT
|
v
SHIP SAFELY --> WATCH THE SIGNALS --> ITERATE
Move fast when exploring; slow down where correctness, data, and reliability matter.
Open-source collaboration, hackathons, AI experiments, practical product ideas — reach out.
WITH GREAT POWER COMES GREAT DEBUGGING.
Independent Spider-Man-inspired fan theme. Not an official Marvel product.