Full-stack · Backend · GenAI
I build products end to end: backends and data pipelines, LLM features with guardrails around them, and the web or extension interfaces people use. I go deep where it matters, down to a database engine or a wire protocol. Each project below links to its code and says what it doesn't do yet.
Bachelor's in Computer Science · BITS Pilani · 8.82 CGPA · 2024–2028
Updated
Apache Superset #43985: Vary SQL Lab ad-hoc query cache key by effective RLS predicates. Open and approved by a reviewer; not merged yet. Pull request
Apache Superset #44248: Fall back to the lean Dockerfile target when publishing a pre-#44100 release. Open and approved by a reviewer; not merged yet. Pull request
Vitest #9662: Add mergeTests utility to compose TestAPI fixtures. Open. The maintainer has requested changes to the tests. Pull request
LEXIS, a retrieval-augmented generation (RAG) system that is still a work in progress. Recent work replaces a placeholder entailment check with a real model and wires in cross-encoder reranking, which is off by default. Repository
OSS Hunter, an engine that scores a repository's issues for complexity with an LLM and ranks them by graph-propagated importance. Early stage, and the repository is private for now.
FlashFlow, a Go lab that replays routing policies on identical traffic to show why one collapses under load. Project page →
A security and honesty pass over CampusSync, VaultTabs, RecoveryOS and AgentBrake, including AgentBrake's fail-open bugs. The AgentBrake record →
The 5 projects with the most verifiable engineering. Each page shows the evidence, links to the source files, and says what isn't done.
Recovers failed payments: an LLM may recommend, but only a deterministic policy engine may act.
Razorpay Buildathon, Track 03 · Aug–Sep 2026
A relational database built from scratch: B+ tree, buffer pool, SQL, optimizer, locking and crash recovery.
Course capstone (two people) · Jun 2026
Semantic bookmark manager: pgvector search, a worker pipeline, and recommendations gated by regression tests.
Personal project, ongoing · Jul 2025 – present · 455 commits
A browser-based debugger for Server-Sent Events: query language, replay, and multi-tab stream sharing.
Personal project · Feb–Mar 2026
A Go lab that replays routing policies on identical traffic and classifies why one of them collapses under load.
Personal project · Aug–Sep 2026 · 226 commits on 5 days
10 merged pull requests to projects I don't own: 7 in Apache Superset, 1 in Appwrite and 2 in Vitest. Mostly bug fixes with tests, and each was reviewed and merged by a maintainer.
Made EncryptedType resolve SECRET_KEY lazily, fixing silent failures of the re-encrypt-secrets command during key rotation. Added an integration test.
Added migrations and security-manager logic that remove deprecated permissions left behind by upgrades, with tests and upgrade notes.
Fixed permission lookups for names like db.dbo.table on engines without catalog support, across the DB engine specs and the security manager.
Dynamic native-filter option queries now use the filter-state cache timeout, fixing stale dropdown values.
Added a filter-level display format for native time filters, reusing the D3 time-format setup.
Masked database passwords when metadata database URIs are printed in startup warnings and by test-db.
Stopped keyboard events bubbling to the menu handler so arrow, Home and End keys work in those inputs.
MFA recovery codes were rejected as invalid because a type constant was lower-cased. Fixed it and added an end-to-end test. First released in v1.9.0-rc.1 (not in 1.8.1).
A test.only nested inside describe.only now runs only the .only tests, matching Mocha. Added a regression test. First released in v4.0.17.
Detects when tsc prints its help text (no tsconfig found) and raises a clear error instead of dumping over a hundred lines. With tests.
Open now
I keep the misses in the record rather than tidying them away. Two examples; the decisions and investigations pages have the rest.
In AgentBrake I wrote a circuit-breaker policy with open, half-open and reset logic, and unit-tested it. Then I found that nothing in the proxy ever reports a failure to it, because the child's output is piped straight through. I fixed it months later by parsing the server's responses. The lesson: wire the signal before writing the policy.
Read the full recordIn MiniDB I expected the vectorized executor to beat the Volcano-style one by about ten times. It reached roughly 1.2 to 2.2×, and the benchmark doc explains the overhead instead of hiding the miss. The lesson: measure first, and publish the number you got.
Read the full recordFeb 19, 2026 · 10 min read
What building a small utility taught me about the difference between algorithmic complexity and integration complexity and why the second one is harder.
Feb 15, 2026 · 4 min read
Recently, I ran into a strange bug where emojis like 👍 were showing up as corrupted characters in stack traces. This is a deep dive into how JavaScript's atob() handles Base64, Unicode, and the difference between binary strings and text.
The rest of the site is a small control-plane for my engineering notes. Press Ctrl/⌘ K to search it.