Mathi

Full Stack developer · Coimbatore · writing since 26 Oct
Day 23 of 30Open to roles

Paid for thinking · 1 why picked at Proof

Picked logs · 1
16 Nov
Picked · 16 Novowned a misschanged the plan
… This got me thinking about how to validate extracted values instead of blindly trusting an AI-generated response. AI can help extract the information, but the application still needs to handle validation, missing fields, and incorrect outputs.
Under a month on Proof
■ weeks with logs■ quiet week
29 logs · 12 say why
How Mathi thinks
Decides and owns the call — moves before being told.
↳the log that shows it
articulate·curious·decisive
How Mathi thinks, mapped · lit by their own logs · 6 showing
How they reason · the four moves
Evidence
argue from what you saw, not what you assumed
Alternative
name the option you didn't take
Assumption
surface what you're taking on faith
What their logs say
Curious
you ask, not just assert
Articulate
you write the full why
Ownership
you own the call & the miss
Showing — it’s in their logs; tap to see them allno %, no rank — evidence, not a score.
About

Lead Full Stack Engineer with 5.5 years building SaaS and AI-powered products for US clients, promoted from QA to team lead in five years. Most recently led 4 engineers and acted as the technical bridge between US stakeholders and the India delivery team.

TypeScriptPythonRubyReactNext.jsReact NativeRailsFastapiPostgresSupabaseDockerKafkaLLMs / GenAIOpenaiAnthropicRagRest ApisTestingTddJavaScript
🎓 M.Sc. Mathematics · Bharathiar University · 2018
↗ mathi1322.github.io
Careerself-reported
Jan 2026 – Jul 2026
Lead Full Stack Developer
Drylogic Solutions · Coimbatore
Jan 2025 – Dec 2025
Senior Full Stack Engineer
Drylogic Solutions · Coimbatore
Apr 2022 – Dec 2024
Full Stack Engineer
Drylogic Solutions · Coimbatore
May 2021 – Mar 2022
QA Engineer, manual & automation
Drylogic Solutions · Coimbatore
Feb 2021 – Apr 2021
Trainee
Drylogic Solutions · Coimbatore
Projects
GitHub
GHgithub.com/mathi1322 ↗

Repos show when they’re linked from a log. Proof doesn’t count stars or commits.

Mathi · the character an interview can’t show

29 logs over 4 weeks — read from their own words.

Character · what an interview can’t reach
Mathi has written 29 logs so far — not enough yet to read a way of working from. The logs themselves are the thing to read.
The judgment behind it ↓
And the technical judgment holds up
Evidence
5
Alternative
6
Assumption
5
Tradeoff
0
Showed up · 21 days
0 stuck + 0 hard days, logged anyway. Nobody farms honest bad days — that’s why this counts and a streak doesn’t.
Read Mathi’s 29 logs ↓sign in to connect with them
Their logs29
Mathi SFull Stack developer20hLeads BrewCode Tech →
Mathi SFull Stack developer1dLeads BrewCode Tech →
Mathi SFull Stack developer3dLeads BrewCode Tech →
Mathi SFull Stack developer3dLeads BrewCode Tech →
Whybecause, different ERP systems can have different processing times and availability.Using a message-driven approach allows us to publish integration requests and let consumers process them independently. So, we used Kafka to also track processing states and failures without blocking the original user workflow, like... publishing, consuming, retries, error handling, and reconciliation.
Mathi SFull Stack developer4dLeads BrewCode Tech →
Mathi SFull Stack developer5dLeads BrewCode Tech →
Whybecause, the important part of a migration is keeping the existing product behavior while adapting to the new API model.So, I prefered understanding the full integration lifecycle before changing the implementation. Otherwise, we risk fixing the API call while breaking the workflow around it.
Mathi SFull Stack developer6dLeads BrewCode Tech →
Whybecause...one thing I noticed is that the developers sometimes focus on optimizing application code while the actual bottleneck lies in the database. A query might look simple but become expensive when it joins multiple tables, processes large datasets, or executes repeatedly for different dashboard components. I don't think adding indexes everywhere is a good optimization strategy either. Indexes have their own storage and costs. My approach is to understand the query first, inspect its execution plan, identify expensive operations, and then consider indexing, query restructuring, or reducing unnecessary data fetching. The important part is identifying the actual bottleneck before deciding what to change.
Mathi SFull Stack developer9 NovLeads BrewCode Tech →
WhyWhile working on a business review platform, I integrated Google Maps APIs to fetch information about places.One important thing was the cost of repeated requests to get the same information. So, I introduced a caching layer to reduce repeated requests and help control API usage costs. But, How long should the data remain valid? What happens when information changes? Which data can be reused safely? The interesting part is finding the balance between data freshness, response time, and external API costs. So, I implemented a layer to store the data whenever we fetches from external API with 30 days TTL and a background job to clear old data, so user can get the updated data accordingly. A good caching strategy should solve the actual usage pattern, not just cache everything blindly.
Mathi SFull Stack developer7 NovLeads BrewCode Tech →
Picked at Proof · 16 Nov · owned a miss · changed the plan
Whybecause document extraction can be unpredictable. Different invoice layouts, scanned documents, and inconsistent formatting can affect the results. I don't want a temporary failure with one provider to become a complete failure. This got me thinking about how to validate extracted values instead of blindly trusting an AI-generated response. AI can help extract the information, but the application still needs to handle validation, missing fields, and incorrect outputs.
Mathi SFull Stack developer7 NovLeads BrewCode Tech →
WhybecauseOne thing I've learned is that rewriting code isn't necessarily an improvement. Existing applications contain business rules and edge cases that aren't documented anywhere. Before changing the code, I need to understand what the existing code does and what depends on it. The goal isn't just a cleaner code. It's making the application easier to maintain without accidentally changing how the product works.
Mathi SFull Stack developerDecided · 6 NovLeads BrewCode Tech →
Mathi SFull Stack developer5 NovLeads BrewCode Tech →
Mathi SFull Stack developer4 NovLeads BrewCode Tech →
Mathi SFull Stack developer3 NovLeads BrewCode Tech →
Mathi SFull Stack developer2 NovLeads BrewCode Tech →
Mathi SFull Stack developer1 NovLeads BrewCode Tech →
Mathi SFull Stack developer31 OctLeads BrewCode Tech →
Mathi SFull Stack developerLearned · 30 OctLeads BrewCode Tech →
Mathi SFull Stack developer29 OctLeads BrewCode Tech →
Mathi SFull Stack developer29 OctLeads BrewCode Tech →
Mathi SFull Stack developerBuilt · 28 OctLeads BrewCode Tech →
Mathi SFull Stack developer28 OctLeads BrewCode Tech →
Mathi SFull Stack developer27 OctLeads BrewCode Tech →
Free write27 Oct 2026
A mock interview with Radhakrishnan Selvaraj. What did you learn about yourself in there? It was a good experience, Thanks to RK. I got to know where I was lacking and how should I be prepared. Even though I know how to get things done, RK pointed out what is important that an interviewer expects from us and it was really helpful.
Mathi SFull Stack developerAssumed · 26 OctLeads BrewCode Tech →
Mathi SFull Stack developerBuilt · 26 OctLeads BrewCode Tech →
Mathi SFull Stack developerLearned · 25 OctLeads BrewCode Tech →
◈ Open to sharpen
Mathi SFull Stack developerBuilt · 25 OctLeads BrewCode Tech →
Mathi SFull Stack developerBuilt · 25 OctLeads BrewCode Tech →
Want a page like this?
Write what you tried, a few lines a day. People who’d hire you read it.
Start your Proof