07
← Transmissions / Products

Lakshya V2: Building the Job Search OS I Wish Existed When I Was Hunting

Most job searches are spreadsheets in disguise. Lakshya is an attempt to turn the hunt into a system — fit scoring, scraping, and cover letters that don't sound like GPT wrote them.

The modern job search is completely broken. Despite living in the golden age of software, finding a technical role usually devolves into a disjointed nightmare of managing thirty different browser tabs, keeping track of manual applications in an Excel spreadsheet, and constantly re-typing the same cover letter over and over again.

I wanted to stop tracking my job search and start engineering it. That’s the genesis of Lakshya V2—not an application tracker, but a true Job Search OS.

Systematizing the Hunt

The goal was absolute automation with zero loss of quality. I built the architecture aggressively around Next.js and Supabase, establishing a fast, reliable database layer that actually felt like an application and not a spreadsheet.

But the real magic happens in the data ingestion pipeline.

I integrated the Apify, building a direct scraper that constantly monitored LinkedIn for highly specific roles. But I didn't want every job. I wanted the right job. I built an uncompromising "India-first" geographic filtering mechanism that immediately discarded irrelevant noise, leaving only hyper-targeted results.

The Fit Scoring Engine

Just having job listings wasn't enough. I needed the system to evaluate the jobs for me.

Whenever the Apify worker drops a new listing into Supabase, a secondary background trigger fires up the Fit Scoring Engine. This takes the raw text of the job description and passes it against my own hard-coded resume data matrix using a combination of heuristics and LLM APIs.

If the score returns higher than an 85% match, the system automatically tags the role as "High Priority". Only then does it execute the final payload.

Claude API: Killing the Robot Voice

If there is one thing hiring managers despise today, it is cover letters that sound like they were generated by ChatGPT. The prose is always florid, bloated, and entirely inauthentic.

To fix this, I completely skipped the standard OpenAI endpoints and integrated the Claude API. By feeding Claude highly-structured, explicitly constrained prompts containing both my exact resume history and the targeted job requirements, it generates cover letters that actually sound like a human engineer wrote them.

Lakshya V2 isn't just a side project; it's a weaponized approach to career mobility.