B.Tech CSE '27 · VIT Vellore · GPA 9.27
Distributed backends that hold under load. Retrieval systems that stay grounded. Deterministic before probabilistic.
01The System
6 systems
02Approach
Engineering principles, not preferences.
- 01
Deterministic before probabilistic
A model should never be the first thing to touch your data. OCR, rules, and entity extraction establish a verified factual baseline first — then inference operates on ground truth instead of guessing at it. Every ungrounded answer is a bug you shipped on purpose.
- 02
Design for the failure, not the demo
The happy path writes itself. What matters is the replay that arrives twice, the message that poisons a partition, the instance that scaled up and lost consensus on a counter. Idempotent writes, dead-letter queues, and atomic state aren't extras — they're the actual system.
- 03
If it isn't measured, it isn't real
A throughput claim without a benchmark is a hope. I instrument the decision path before I tune it — Micrometer on every rate check, OpenTelemetry across every service hop — because you cannot fix a bottleneck you can only speculate about.
- 04
End-to-end ownership
Schema, ingestion pipeline, inference endpoint, and the UI rendering the result. Owning the whole chain is how you find the bottleneck that lives between two layers — the one nobody on either side considers theirs.
03About
I build backend systems that survive real traffic and AI systems that don't make things up. Currently a software engineer intern at Hanco Automotive in Jeddah, shipping microservices for a SaaS platform serving 10,000+ active users. Before that, Techvaria in Bangalore, moving 50K daily events through FastAPI and SQL pipelines.
My work sits across distributed backends, high-throughput data pipelines, and grounded retrieval systems. Kafka and ClickHouse on one side, hybrid BM25 + vector search on the other — and the same instinct underneath both: establish what is verifiably true before you let anything infer.
I think in failure modes. Idempotent writes, dead-letter queues, atomic Redis operations, deterministic baselines before an LLM ever sees the document. Every one of those is a bet about what breaks at 3am — I'd rather make the bet on purpose.
- 9.27
- GPA / 10
- 2
- Engineering internships
- 6
- Systems shipped
05Experience
Education, projects, and work — in order.
Aug 2023
VIT Vellore
B.Tech Computer Science & Engineering
Started CS. Data structures, operating systems, networks, DBMS, system design — and a growing preference for the layers underneath the framework. Currently at a 9.27 GPA, graduating May 2027.
Jun 2025
Techvaria
Software Engineer Intern · Bangalore, IN
First production engineering role. Real data, real stakeholders, real deadlines.
- Implemented FastAPI backend services and SQL/MongoDB data pipelines processing over 50K daily user events for real-time Power BI analytics dashboards.
- Optimised complex database queries using Pandas and NumPy for behavioural analytics — 20% faster queries and 30% less manual reporting effort.
Late 2025
FlowSight · FileLens
Full-Stack & Systems Programming
Built an OCR-driven financial analytics platform on Spring Boot and React, then went a layer lower — a Rust and Tauri 2 desktop app doing multithreaded BLAKE3 deduplication across 100GB+ datasets. Learning where the abstraction actually costs you.
Early 2026
ClearClause · Traya · FluxLimit
Applied AI & Distributed Systems
Three systems in parallel: a deterministic-first RAG platform with hybrid BM25 + vector retrieval, a Kafka-and-ClickHouse observability pipeline that survived 115K requests in 50 seconds without losing a record, and a Redis-backed rate limiter benchmarked at 4.9M checks per second.
May 2026
Hanco Automotive
Software Engineer Intern · Jeddah, KSA
Backend engineering for a SaaS platform in production.
- Engineered backend microservices and PostgreSQL database features for a platform serving 10,000+ active users, integrating 40+ JUnit tests to reach 85% code coverage and validate core REST APIs.
- Automated containerised deployments with Docker and GitHub Actions CI/CD, cutting manual deployment overhead by 40% while delivering production software across 3 Agile teams.
2026 →
Now
Jeddah · Oracle & AWS certified
Shipping at Hanco, deepening distributed systems work, and closing the gap between retrieval quality and answer grounding. Open to SDE and AI engineering roles for 2027.
06Now
What’s on the workbench — August 2026
Building
Backend microservices at Hanco Automotive
PostgreSQL features and REST APIs for a SaaS platform with 10,000+ active users — 85% coverage, containerised deploys, three Agile teams.
Exploring
Retrieval grounding beyond RRF
Where hybrid BM25 + vector fusion still misses, and how much of that is rerank quality versus chunking strategy. Answers I don't have yet.
Reading
Designing Data-Intensive Applications
Kleppmann, chapter 9 — Consistency and Consensus. Rereading it after Traya hit the exact problems it describes.
Last shipped
Traya — distributed observability
Kafka to ClickHouse telemetry pipeline across 8+ services. 115K requests in 50 seconds, 0% data loss under stress.
07Contact
Graduating May 2027 and open to SDE and AI engineering roles. If you’re working on something genuinely hard — distributed backends, high-throughput pipelines, retrieval that has to stay grounded — I want to hear about it.