suryansh.Full-stack engineer · AI & platforms
Let’s talk
Open to the right opportunityPune, India /IST

Suryansh Singh · Software engineer

Hi, I’mSuryansh.

I build production AI products, backend services, and developer platforms. I like working on both the application code and the systems that keep it running.

Explore my workA little about me

Dot-grid hero animation with four modes: a full-stack request round trip, a CI/CD rollout across a cluster with self-healing, a line of code traced through HTTP, TCP and bits, and a text prompt generating an image.

A request, from interface to database and back.CLICK TO SEND A REQUEST
See the projectsProjects & engineering notes

01 / Selected projects

A few things I’ve
put into the world.

Products and systems I’ve built, the decisions behind them, and the outcomes so far.

View all projects
01 / End-to-end product engineeringProfessional workIn production

A complete call-quality review platform

From call scores
to human review.

Turned disconnected AI call scores into a complete review experience: find the conversations that need attention, review the evidence, and make sure the right action follows.

One complete workflowfrom automated flagging to human review and follow-up
Python / FastAPI / React / TypeScript01 02
02 / AI developer toolingIndependent projectActive development

TokenMuncher

Finding source code
for coding agents.

A personal project exploring how coding agents find the code they need. Resolve a named symbol and assemble useful context within an explicit budget.

~54%fewer candidate source tokensSmall mixed-language fixture; compared with name-based file dumps.
Python / Tree-sitter / SQLite / AI02 02

Also in the collection

More work, in brief.

03 / Production AI systems

Customer care that knows when to hand over

Built a production AI care agent that verifies order status against live data and hands other requests to Care or Sales with clear context. One codebase serves four markets.

4 markets verified order lookup and structured human handoffsRead the case study
04 / AI product engineering

Customer intelligence with evidence

Built a conversational analytics application that helps people investigate customer experiences—and trace each answer back to the data behind it.

~150k customer conversations made queryableRead the case study
Explore all 8 projects

02 / A bit about how I work

How I tend to work.

I usually start by finding out where someone gets stuck. Then I work through the product and system changes needed to fix it.

01

Start with the workflow.

Before deciding what to build, I want to understand what someone is trying to do, what gets in their way, and what they do by hand today.

02

Try the uncertain part first.

If I’m unsure an approach will work, I’ll try a small implementation before building the rest around it.

03

Stay involved after launch.

I like working on what real use reveals: a confusing step, a failed request, or something the team still has to fix manually.

A little more about me

03 / Recent notes

What I’ve been
learning lately.

Notes on things I’m trying to understand. Some come from projects; others start with something I’ve read.

01 / Database Systems

Why PostgreSQL Keeps the Row You Updated

Follow one UPDATE through heap pages, row versions, snapshots, HOT, and VACUUM to see why the old value cannot disappear immediately.

Read the note
02 / Distributed Systems

When Replicas Disagree: Resolving Inconsistency in Distributed Key-Value Stores

Follow a branching record through vector clocks, reconciliation, and last writer wins—and see how Dynamo and DynamoDB make different choices.

Read the note
03 / AI Engineering

An answer is only as good as its denominator.

What building the customer-insights platform taught me about the distance between a convincing AI answer and a defensible one.

Read the note
From the three most recently updated topics.View all notes