Hey, I’m Jitendra.
I’m a frontend dev for whom details matter.
I hate slow stuff.
I hate repetitive work even more when it can be automated.
Here’s what I do about it.
I’ve spent the last few years building products, fixing frontend stuff that annoyed me, and automating work I got tired of doing manually.
Somehow all of it connects.
This is the map.
- 15 channelsPeople used them where they already worked.
- 600+ questionsCode, analytics, product, incidents. Basically whatever got thrown at them.
- Specialists > one giant botDifferent jobs, different context, different tools.
- Jira → branch → PRSome requests go all the way to something ready for review.
- 110+ merged PRs
- Everything in one placeSessions, tickets, jobs, incidents, releases.
- Mistakes should teach the systemHuman corrections shouldn’t disappear into Slack history.
- Propose → review → approve → shipShared behaviour changes need an actual workflow.
- Realtime voiceNot a form pretending to be a conversation.
- LiveKit + React
- People actually used itAlways preferable to a shiny demo nobody opens twice.
- Rules > LLMIf the same input should give the same answer, I don’t want the model freelancing.
- Config-drivenProgression and constraints stay predictable.
- AI for messy inputsResearch and interpretation are where it actually helps.
- 3D body mapEasier than reading recovery off a muscle spreadsheet like a psychopath.
- The model reasons. The system remembers.
- Structured memoryPeople, projects, preferences, commitments.
- LLM ≠ databaseMemory belongs in actual storage, not a giant prompt.
- Brain ↔ handsThink remotely. Act locally on the phone.
- Shared app chunk −60%First-load CSS came down 81% on top of it.
- 404 dead modules removedAnd 88 MB off the static deploy payload.
- Measure firstOtherwise you’re just moving code around with confidence.
- Trace → remove → split → verify
Not everything can load instantly. It still shouldn’t feel slow.
Then somebody let me build a game.
360×640 → 8K
- 3.3K+ monthly filtering partners
- 85K+ shortlists
- Search is a productThe hard part wasn’t the search box. It was everything around it.
So that’s roughly how we got here.
Frontend is still home base.
But somewhere along the way I started caring just as much about what sits underneath the interface: architecture, performance, product decisions, automation, agents and the systems around them.
Turns out I mostly like making annoying things less annoying.
Patterns I keep repeating
Details matter
Loading states count. Empty states count. Weird edge cases count.
If somebody sees it, it’s part of the product.
Fast means fast
Sometimes that means shipping less JS. Sometimes it means making unavoidable waiting feel invisible.
Either way, “it loads eventually” is not a performance strategy.
Fix the system once
If I’m solving the same problem for the third time, I probably shouldn’t be solving it there anymore.
Shared infra, reusable patterns, automation. Less repeated nonsense.
Automate boring work
If people are repeating the same steps every day, software should probably be embarrassed.
AI when it earns its place
Use models for fuzzy reasoning. Use rules when consistency matters. Use databases for memory.
Stop asking one thing to do every job.
Own the problem, not the rectangle
I’m frontend-heavy. I’m just not precious about where frontend ends.
If the solution needs product thinking, APIs, analytics or automation, cool. That’s still the problem.
Things I’ve been thinking about
Occasionally a rabbit hole survives long enough to become a LinkedIn post.
Usually AI, engineering, memory systems, product decisions or something I built and then overthought.
The stuff I keep ending up doing
Frontend
This is still home base.
- React / Next / TypeScript
- Architecture & performance
- Design systems
- Complex + realtime UI
- Large data interfaces
Product
I usually wander past the ticket.
- Flows & state
- Analytics
- Technical trade-offs
- Shipping and iterating
- Working out what the thing should actually do
AI / Automation
Mostly the useful kind.
- Agents & MCP
- Tool use
- Workflow automation
- Structured memory
- Human review in the loop
Across the stack
Enough to not stop at the browser.
- Node / Nest
- REST APIs
- Postgres & Prisma
- Mongo / Firebase
I’m frontend-heavy. The rest mostly happened because products stubbornly refused to end at the browser boundary.
I’m mostly looking at frontend-heavy product roles, frontend/platform work, and interesting AI products.
I’m also open to building with small teams when the problem is worth getting distracted by.
If it involves:
- complicated UI
- performance
- ugly workflows
- repetitive work
- internal tooling
- AI that needs to actually do something
If that sounds remotely like what you’re working on, you know where the email button is.