bbsr --:-- you --:--
scroll 000%
open to work
// anshu.config.ts
const anshu = new Engineer({
  base:   "bhubaneswar, odisha",
  degree: "b.tech — iter, soa '26",
  stack:  [python, java, react, sql, rust],
  builds: ["ai apps", "pipelines", "automation"],
  status: OPEN_TO_WORK,
});

? want to watch a prompt get routed

Anshuman

Tripathy ଅଂଶୁମାନ — “the radiant one” software engineer
ai · data · automation
( about )move the cursor over the sun ↑

I build software that decides where work should go router.py which model answers a prompt, which path an image request takes, which table a log line lands in. Then I automate it jenkins so nobody does it by hand again.

route, don't guess

Right model, every prompt.

My chatbot aggregator sends each prompt to GPT-4o, Claude, LLaMA, DeepSeek or Meta AI based on intent, model strengths and context.

automate the daily

If it runs daily, it runs itself.

Jenkins-scheduled data pulls, ETL, model training and performance reports — with minimal manual work.

fail over, don't fall over

Plan for the API going down.

When hosted image models are unavailable or rate-limited, local Stable Diffusion takes the request.

Selected work

( 03 projects + archive )
drag or scroll →
archivegithub.com/Anshuu0046
visit github ↗
04

Everything else, on GitHub

ongoing
01 / 04
click a card to open the case study
( capabilities )click a row to expand

What I build

Assistants that pick the right model for each prompt, and image tools that keep working when the API doesn't.

  • GPT-4o · Claude · LLaMA · DeepSeek · Meta AI01 router
  • Cohere API — summaries, QA, code, writing01 router
  • Hugging Face API02 failover
  • Stable Diffusion, local inference02 failover

Datasets collected, cleaned, validated and loaded every day — without anyone babysitting them.

  • Daily collect → preprocess → validate03 pipeline
  • XML logs → tables → data warehouse03 pipeline
  • Performance reports, generated automatically03 pipeline
  • SQL · database managementcoursework

Scheduled jobs and scripts that take repetitive work off people's plates.

  • Jenkins — scheduled pulls, training jobs, reports03 pipeline
  • Python automation scripts03 pipeline
  • PowerShell tooling03 pipeline
  • Linuxdaily driver

Dashboards teams actually read, and a Chrome extension for talking to five models at once.

  • Interactive dashboards + daily summaries03 pipeline
  • Chrome-extension chat front end01 router
  • React · Tailwindcore stack

Data structures, algorithms and systems programming — the base everything else stands on.

  • Java · C · C++ · Rustlanguages
  • Google Cloud Platformcloud
  • GitHub · VS Code · Eclipsetools
  • Communication · collaboration · problem solvinghow I work
( log )experience + education

The record

2025.04 — 2025.07

Unified Mentor

Data Science Intern internship
  • Built a daily pipeline to collect, preprocess and validate datasets — less manual overhead, faster model cycles.
  • Python scripts and PowerShell tools that turn structured XML logs into tables and load them into data warehouses.
  • Jenkins-automated ETL: scheduled data pulls, triggered model training, generated performance reports.
  • Interactive dashboards and daily summary reports in HTML, CSS and JavaScript for cross-functional teams.
2023.09 — 2026.06

ITER, SOA

Bachelor of Technology bhubaneswar, odisha

relevant coursework

data structuresalgorithms analysisdatabase managementartificial intelligencesoftware methodologyinternet technologysystems programmingcomputer architecture
( lab )a small instrument

Play the sun

Eight keys tuned to raga Bhupali — a five-note scale from Indian classical music. Each note throws a ray.

type a s d f g h j k, or tap the keys · sound is

( contact )open to software, ai and data roles

Contact

write to

swayamshree16@gmail.com
linkedin ↗ github ↗