Tejas Lokhande

Software engineer · curious about how things work, persistent when they don't

--:-- in Pune, India

Open to SWE roles GitHub Email

The 30-second rundown

Now: Shipping bug fixes to Infisical, the open-source secrets manager: KMS, SCIM and webhook fixes in its TypeScript backend.
Builds: Backend-heavy systems: reconciliation engines, CQRS services in Go, transactional booking systems on PostgreSQL. I like the part where the data is messy.
Stack: TypeScript, Go, Node.js, Next.js, React, PostgreSQL, MongoDB, Redis, Docker, AWS and Azure.
Looking for: Software engineering roles where I get to own real systems end to end.

Experience

Infisical

2026 · Open source · Remote

Open-source contributor · Backend & frontend

Infisical is an open-source platform for secrets, certificates and privileged access. I pick up bugs from its issue tracker, trace them through the codebase and send fixes upstream.

TypeScriptNode.jsPostgreSQLReactSCIM

Selected work

Live demos are usually switched off. These run on real cloud instances (AWS, Azure, managed databases) that cost real money to keep up, so I power them on for interviews. The code is always public, and I'm happy to bring any of them up for a call. Ask for a demo →

Applied AI · Fintech

payment-reconciliation-engine code

Built for the Razorpay AI Buildathon (AI Finance Controller track). Reconciles 920 payment records across three deliberately inconsistent sources: 65.22% matched with 0 false positives, measured against an answer key the engine can't read. The model decides nothing; the match rate is identical with the LLM dead, live or cached.

TypeScript · Next.js · PostgreSQL · LLM tool-calling

Instance off · watch the 5-min pitch

reconcile --report

RECORDS

920 across 3 sources

MATCHED

65.22% of a 93% ceiling

FALSE POSITIVES

0 ✓

EXCEPTIONS

212, each classified, explained and priced

Distributed systems · Cloud

go-url-shortener code

A URL shortener split CQRS-style: redirects vastly outnumber creations and have a tighter latency budget, so reads and writes are separate Go services sharing one MongoDB Atlas instance. Hexagonal architecture, infrastructure as code, and keyless OIDC-based CI/CD into Azure.

Go · MongoDB Atlas · Azure · GitHub Actions · OIDC

Instance off · live on request

architecture

POST /shorten→writer

GET /:code→reader

↓

MongoDB Atlas

reads scale independently of writes

Full-stack · Data integrity

invitrack code

Inventory and booking management with RBAC, OAuth and audit logging. Double bookings are prevented with PostgreSQL transaction-scoped advisory locks, not Redis; inventory edits use optimistic locking and reject stale writes with a 409. Audit events go through BullMQ with retries and a dead-letter queue.

TypeScript · Express · PostgreSQL · Prisma · Redis · BullMQ · React · AWS

Instance off · live on request

PATCH /inventory/items/42

REQUEST

{ "qty": 8, "version": 3 }

CURRENT

version 4 (someone got there first)

RESPONSE

409 Conflict

stale writes surface, never silently overwrite

Everything else on GitHub →

Off the clock

Currently reading

Failing Forward

John C. Maxwell

A book about treating mistakes as stepping stones rather than verdicts. It fits how I debug, too.

Gym

Regular training. Progressive overload works on code as well as on squats.

Reading

Mostly books on growth, habits and how people think.

Guitar

Learning songs one chord change at a time. The best way to switch off from a screen.

Connect

GitHub Email Get in touch