Abhimoy Mukherjee Let’s talk ↗
OPEN TO INTERNSHIPSBENGALURU, INDIA  ·  THINKING BEYOND
Abhimoy in a black hoodie, looking toward the sunlight

BACKEND ENGINEERING  ·  DEVOPS  ·  CLOUD

Abhimoy Mukherjee

Learning by building.

Third-year Information Science & Engineering student exploring backend systems, cloud infrastructure, and DevOps through hands-on projects.

Explore my work ↓Start a conversation ↗
SCROLL TO DISCOVER  ↓
9.40CGPA / 10.00
200+DSA problems solved
01 / THE PERSONBEHIND THE SYSTEMS

CURIOSITY → ENGINEERING

Not just writing code.
Understanding how
systems work.

I’m a third-year Information Science & Engineering student at Dayananda Sagar College of Engineering in Bengaluru.

I’m drawn to what happens beneath the interface: APIs, data, authentication, business logic and the decisions that make software behave predictably. That curiosity has led me to explore backend architecture, cloud engineering, DevOps and reliable deployment.

I like building things, finding out why they work, and tracing what happens when they don’t. Correctness, security and reliability are the questions I keep coming back to.

A little more about me ↗
02 / SELECTED WORKBUILT TO BE USEFUL

Systems with
something at stake.

Projects that have made me think harder about money movement, trust boundaries and what a system should do when it isn’t sure.

01 — FINTECH / RECONCILIATIONRAZORPAY BUILDATHON
INPUT RECORDSPaymentsSettlementsBank entries
RMATCH ENGINEDeterministic
firstRules → validation
HIGH CONFIDENCEMatched
AMBIGUOUSHuman review
AICLASSIFY & EXPLAIN
RECONCILIATION PIPELINE 01 / 03

PAYMENT RECORDS, MADE LEGIBLE

ReconAI

Payment, settlement and bank records don’t always line up cleanly. ReconAI matches the clear cases with deterministic rules and uses AI to classify and explain the genuinely ambiguous ones.

Ambiguous cases go to human review. The evaluation used generated records, 12 hand-built held-out cases and 32 Razorpay test-mode payment anchors; settlement and bank records were synthetic.

1,168deterministic matches
at 100% precision
44ambiguous cases
routed for review
1,212records in documented
evaluation
PYTHON · FASTAPI · POSTGRESQL · SQLALCHEMY · GROQ · DOCKER
Repository ↗Live demo ↗
Explore the engineering decisions ＋

THE PROBLEM

Reconciliation must account for missing settlements, duplicates, delays and fee differences. Sending every record to a language model would add uncertainty where explicit rules can do the job.

THE APPROACH

Normalize records, apply identifier, amount, date-tolerance and fee-adjustment rules first. Reserve the LLM for bounded classification and explanations, then validate structured outputs.

WHAT I LEARNED

Make uncertainty visible. Escalate ambiguous cases rather than silently deciding financial outcomes. The repository reports an 85.73% overall match rate; split payouts remain a known limitation.

02 — FINTECH / AI SECURITYPROTOTYPE · THREE-PERSON TEAM
TRUST BOUNDARY / LIVE TRACE
TRUSTEDPayment intentLocked plan01
CAPABILITY
GATE
UNTRUSTEDPDF instructionVendor email02
QUARANTINE LLM
NO TOOL ACCESS

KEEPING UNTRUSTED DATA IN ITS LANE

PayShield Arena

A prototype for exploring prompt injection in AI-powered payment agents. It separates a locked trusted plan from untrusted external information, then gates execution through a capability-tagged interpreter.

Hackathon-scale prototype with a mock SQLite ledger; it is not connected to real payment rails or deployed as a commercial security product.

3 / 3tested injection vectors
blocked in evaluation
0false positives observed
in tested legitimate cases
FASTAPI · GROQ · SQLITE · REACT · SSE · RENDER

My contribution: CaMeL-inspired pipeline and Arena UI.

Repository ↗
Explore the engineering decisions ＋

THE PROBLEM

External documents and messages can contain instructions that try to redirect a payment agent away from user intent.

THE APPROACH

A planner locks the intended action before external data is read. A taint-aware interpreter enforces capabilities; a quarantined LLM has no tool access. SSE traces show how trust labels move through the pipeline.

THE EVALUATION

Malicious PDF, poisoned email and tampered vendor-record injections were blocked in the tested scenarios. The baseline agent was compromised in all three. These scoped prototype results do not establish production protection.

03 — CLOUD / DEVOPSCONFIDENCE-GATED REMEDIATION
01 / SIGNALCI failsGitHub Actions
→
02 / DIAGNOSEClassifyLogs → evidence
→
HIGH CONFIDENCEPull request
OTHER CASESHuman review
AWS · BEDROCK · LAMBDA · API GATEWAY

FROM FAILED PIPELINE TO EVIDENCE

Deploy Doctor

An AI-assisted workflow that classifies deployment failures into six categories and grounds each diagnosis in specific log evidence. A narrow safe-fix path can open a pull request; other cases are escalated for human review.

Automation is bounded: it proposes changes for a limited high-confidence path and does not auto-merge or perform arbitrary remediation.

PYTHON · AMAZON BEDROCK · AWS LAMBDA · API GATEWAY · AMPLIFY · GITHUB ACTIONS

Built to keep a person in the loop when the signal is unclear.

Repository ↗
Explore the engineering decisions ＋

THE PROBLEM

CI/CD failures interrupt delivery, and useful evidence can be buried in logs.

THE APPROACH

GitHub Actions triggers a diagnosis workflow across AWS services. The system classifies failure types, cites log evidence and gates proposed fixes on confidence.

THE TRADE-OFF

Opening a pull request is a reviewable suggestion. Escalating other cases through GitHub issues keeps uncertain fixes under human control.

MORE EXPLORATIONSSMALLER SYSTEMS, SAME CURIOSITY
01 / AI · BACKEND

CVpulse

Resume parsing and job-description matching with structured outputs.

FASTAPI · PYDANTIC · GROQ · DOCKER↗02 / FULL-STACK

Milan

Event RSVP and management with JWT authentication and relational persistence.

FASTAPI · SQLALCHEMY · SQLITE · JAVASCRIPT↗03 / SYSTEMS · BASH

Health Monitor

A command-line utility for CPU, memory and disk thresholds, with logging.

BASH · LINUX · OPERATIONS↗
03 / HOW I BUILDQUESTIONS BEFORE ANSWERS

Good systems begin
with good questions.

Not a formal methodology. Just a way to keep the important things in view as an idea becomes a working system.

01 / 05

Understand

Start with the outcome and constraints. Ask what can fail, who needs to know, and what the system should do when the answer isn’t clear.

IN PRACTICE · RECONAI → AMBIGUOUS CASES GO TO REVIEW
04 / ENGINEERING TOOLKITTOOLS WITH CONTEXT

Turning ideas into
working systems.

A growing toolkit shaped by projects and study. Hands-on experience varies by area; these aren’t proficiency scores.

01 / LANGUAGES

Build & reason

Java 2+ years · DSA

Python

C · C++

02 / BACKEND

Services & data

FastAPI · REST

SQLAlchemy · Pydantic

JWT authentication

03 / DATABASES

Persistence

PostgreSQL

SQLite · SQL

Relational modelling

04 / CLOUD & DEVOPS

Ship & operate

AWS Bedrock · Lambda

API Gateway · Amplify · IAM

Docker · Compose · GitHub Actions

CI/CD · Bash · Linux/Unix

05 / DEVELOPER TOOLS

Daily foundations

Git · GitHub

Linux / Ubuntu

06 / CS FUNDAMENTALS

Under the hood

DSA · OOP

Operating Systems · DBMS

Computer Architecture

05 / THE JOURNEY SO FARSTILL BUILDING. STILL LEARNING.

Progress is a
practice.

Milestones in study, problem-solving and getting closer to the engineering work I want to do.

EDUCATION / 2024—2028
DSCE·

Dayananda Sagar College
of Engineering

B.E. Information Science & Engineering
Bengaluru, India

9.40CGPA / 10.00
Expected graduation · May 2028
AUG 2026

Adobe University Hackathon

Cleared Round 1 DSA assessment; advanced to approximately the top 26,400 teams from ~100,000, then completed a live case study.

JUL 2026

Flipkart GRiD 8.0

Cleared Round 1 profile screening and advanced to a proctored assessment in DSA, SQL and Python/Pandas.

DEC 2024—MAR 2025

Point Blank · DSCE Coding Club

Active member involved in coding-club initiatives.

ONGOING

Practice, one problem at a time

200+ DSA problems solved · 150 active days · Java as my primary practice language.

06 / WHAT’S NEXTOPEN TO INTERNSHIP OPPORTUNITIES

BACKEND · CLOUD · DEVOPS

The next chapter
is still being built.

I’m looking for an internship where I can contribute to real engineering work and learn from experienced teams.

Especially interested in backend engineering, cloud infrastructure, DevOps, CI/CD and platform systems. I’m also preparing for full-time software engineering opportunities in my fourth year (expected graduation: May 2028).

I enjoy connecting with engineers, founders, recruiters and people who care about building useful technology. Always open to thoughtful conversations, mentorship and collaboration.

Discuss an opportunity ↗
BACKEND SYSTEMSCLOUD INFRASTRUCTUREDEVOPS & CI/CDPLATFORM ENGINEERING
07 / LET’S CONNECTBENGALURU, INDIA · 13° N / 77° E

GOOD THINGS BEGIN WITH A CONVERSATION.

Let’s connect.

Recruiter, engineer, or someone building something interesting — my inbox is open.

Your email app will open so you can review and send your message.