2025.11
Atlas, honestly
What a 'connected workspace' actually means when you stop hand-waving over the connector layer.
[field notes]
8 min
40%
faster reporting turnaround
Spacewalk PMO workflows
akhielesh·AI Product Engineer · Vibe Coder
Akhielesh
builds AI products
that hold up
past the demo.
Five years from data to AI products. I ship on weekends and host what I ship — Atlas and DreamStream Studio are the proof.
this site has a terminal. try chat or open atlas.
60%
profiling report time cut
GSK DQ framework
98%
pipeline success rate
GSK ingestion + profiling
16
live connectors on Atlas
Atlas beta
01 / projects
Things I've built and hosted.
A mix of shipped AI products and weekend experiments. The ones I'm proud of make their tradeoffs visible instead of hand-waving over them.
02 / work
Five years of moving
data — and now products.
Day jobs that built the foundation: ETL, profiling, dashboards, automation. Click any row for the receipts.
30%
annual cost savings
Jacobs license analytics
25%
faster intake → delivery
Jacobs Agile Demand
7+
business domains covered
Automated DQ visibility
5 yrs
cross-functional shipping
Research → analytics → AI product
Education
M.S. Information Systems Technology
George Washington University
Washington, DC
Graduate Certificate, Cloud Applications & IT
George Washington University
Washington, DC
B.Tech, Computer Science Engineering
SRM Institute of Science & Technology
Chennai, India
03 / skills
The kit, table-style.
No mastery meters. Just what I actually reach for.
AI product
Connected searchSource-aware actionsStage-based workflowsBYOK model routingPrompt orchestration
Languages
PythonTypeScriptSQLSAS (Base, PROC SQL, macros)JavaScriptBash
Frameworks
React 18/19Next.jsExpressViteTailwindPrismaZustand
Data
PandasNumPyDatabricksAzure Data FactoryETL/ELTStar/SnowflakeSCDReconciliation
Platform
PostgresSupabaseRedisBullMQOpenSearchS3CloudflareRailwayAWS
Analytics
Power BITableauThoughtSpotDashboardsKPI automation
Databases
OracleMySQLPostgreSQLSQL ServerTeradata
04 / writing
Notes from the work.
Field notes from building Atlas, DreamStream Studio, and the data years. These are the working titles — long-form versions are landing soon.
2025.10
ComicForge: queue-backed creativity
Why DreamStream Studio is a job queue with a UI on top, not the other way around.
[engineering]
12 min
2025.08
From ETL to AI product, slowly
Five years of moving data around taught me more about AI products than I expected.
[career]
6 min
05 / now
Right now.
What I'm working on, reading, and avoiding — updated when reality shifts.
last updated · 2026.06
Shipping
Atlas v0.4 — adding source-aware permissions UI and a real audit trail for connector actions.
Building
DreamStream Studio at dreamstreamstudio.ai — hardening the ComicForge pipeline and panel-consistency QC.
Reading
Designing Data-Intensive Applications (again) — different chapters hit different at every stage of building.
Listening
Mostly ambient + tabla. Long focus blocks.
Avoiding
Generic SaaS chrome. Animation for animation's sake. Anything beige.
06 / uses
The setup.
Tools, keys, and pixels that ship the work.
editor
- Cursor
- VS Code
- Vim (for the muscle memory)
languages
- TypeScript
- Python
- SQL
ai
- Claude
- Gemini
- Flux for images
runtime
- Cloudflare Pages
- Railway
- Supabase
- Postgres + Redis
machine
- MacBook Pro M-series
- 27" external
- Vortex Race 3 (Cherry Brown)
desk
- Notebook + Pilot G2 0.5
- One cup of strong coffee · always one
07 / lab
Half-built things.
Side experiments, all public on GitHub. Some will graduate; some are deliberately just vibes.
01 / 04
sourcevault
A data-brain UI — sources, lineage, and downstream impact thought through as one canvas.
02 / 04
elements
The periodic table reimagined as a navigable encyclopedia — element-as-page, deep links, cross-refs.
03 / 04
keystroke-analyzer
Typing rhythm as a soft biometric — dwell, flight, and error patterns visualized and classified.
04 / 04
drone-simulator
A physics + mission-planning sandbox for autonomous drone flight scenarios.
08 / contact
Let's talk.
Best for AI product, full-stack, and platform-y data roles. Also for random ideas at 2am.
tell me
what
you're
building.
site
akhielesh.com
github
linkedin
remarks
status
open to AI product engineering & full-stack roles
akhielesh@portfolio
akhielesh@portfolio — zsh
akhielesh@portfolio:/ $