KASHYAP KEST. 2005MACHINE LEARNING ENGINEER[ CLICK TO SKIP ]
AVAILABLE FOR INTERNSHIPS & COLLABORATIONS

I BUILD MODELS,

SYSTEMS

& PRODUCTS.

~/AI/ML Developer

I'm Kashyap K — a B.Tech undergraduate at PES University, interested in artificial intelligence, machine learning, and building intelligent systems that turn ideas into real-world products.

View Work
Get in touch
Kashyap K
[ PES UNIVERSITY ]BENGALURU · IN
SCROLL
Python/PyTorch/TensorFlow/ConvNeXt/EfficientNet/FastAPI/React/Electron/LLaMa/Kafka/Redis/MySQL/AWS/OpenCV/WebSocket/
[ 01 / ABOUT ]

I build technology that creates real-world impact — and I learn by shipping.

I'm a Computer Science student passionate about Artificial Intelligence, Machine Learning, and building technology that matters. My focus goes beyond learning concepts — I enjoy transforming ideas into projects and real products.

Right now I'm building a strong foundation in AI/ML through deep learning, hands-on projects, hackathons, and continuous self-improvement — with a long-term goal of building intelligent systems that solve meaningful problems at scale.

I believe in learning by building, thinking long-term, and constantly pushing beyond my comfort zone to create things that matter.

Languages

PythonC/C++JavaScriptSQLR

AI / ML

PyTorchTensorFlowConvNeXtEfficientNetLLaMaOpenCV

Systems & Data

KafkaRedisHadoopPySparkAWS

Full-Stack

FastAPIReactFlaskElectronMySQL
[ 02 / SELECTED WORK ]

Things I've built.

02
Multi-Agent · Agentic AI

NexusTravel

A multi-agent system where specialized AI agents collaborate, critique, and negotiate to produce validated travel itineraries. When agents hit a deadlock — fixing the budget breaks the schedule, fixing the schedule re-breaks the budget — a deadlock-detection and compromise-resolution system finds the minimum relaxation needed to converge on a valid plan.

Multi-AgentLLMsPythonNegotiation
MultiAgents
Auto-resolveDeadlock
ValidatedOutput
NexusTravel 1
01
NexusTravel 2
02
03
Systems · Big Data

Distributed Image Pipeline

A distributed image-processing pipeline using Kafka for asynchronous communication between master and worker nodes — enabling scalable, fault-tolerant, parallelized image transformations across the cluster.

Apache KafkaPythonMulti-threadingDistributed
Master–WorkerArchitecture
AsyncComms
Fault-tolerantSystem
Distributed Image Pipeline 1
01
04
Data · Full-Stack

WildTrack

A full-stack MySQL + Flask app to manage and analyze wildlife conservation data — species tracking, habitats, rescue operations, sponsorship funding and medical records — with dynamic dashboards and real SQL logic: triggers, stored procedures, and user-defined functions.

MySQLFlaskSQLDashboards
NormalizedSchema
Triggers + SPSQL
DashboardsView
WildTrack 1
01
WildTrack 2
02
WildTrack 3
03
[ 03 / EXPERIENCE ]

Where I've worked.

JUN 2026 — PRESENT

Product & Innovation Intern

PESU Venture Labs
Bengaluru, India

Working inside PES University's venture studio — helping early-stage startups validate product-market fit and shape deeptech (AI/ML, IoT) ideas through the studio's incubation and acceleration programs.

[ 04 / EDUCATION ]

Where I learned.

2023 — Present

Bachelor of Technology

Computer Science & Engineering
PES University
Bangalore, India
Machine LearningDeep LearningComputer VisionData StructuresAlgorithmsBig DataDistributed Computing
PES University 1
01
PES University 2
02
PES University 3
03
PES University 4
04
PES University 5
05
2021 — 2023

Pre-University (Class XII)

Physics · Chemistry · Mathematics
Vivekananda PU College
Puttur, India
94%Boards
96%ileJEE
3,527KCET
Vivekananda PU College 1
01
2021

Class X (SSLC)

Vivekananda English Medium School
Thenkila, Puttur
96%Class X
Vivekananda English Medium School 1
01
Vivekananda English Medium School 2
02
Vivekananda English Medium School 3
03
Vivekananda English Medium School 4
04
[ 05 / WRITING ]

I write about what I build.

ConvNeXt on 239 Indian Foods

8 min read

Computer Vision /How I trained a ConvNeXt-Small classifier to 87.34% validation accuracy (97.1% top-3) on 156k images across 239 Indian dishes — the vision engine behind NutriVerse.

Reimagining Manufacturing with AI

6 min read

AI Strategy /Notes from PES University’s AI symposium. The lesson that stuck, straight from industry: don’t build models — build decisions that deliver measurable ROI in 6–18 months.

EfficientNet on a Free Kaggle GPU

4 min read

Deep Learning /A hands-on log of fighting CPU crashes and WSL2 CUDA mismatches before Kaggle’s free T4 GPU delivered a ~6× speedup — training 148 food classes over 12 epochs.

Unifying Two Messy Food Datasets

7 min read

Data Engineering /A complete walkthrough of cleaning, fuzzy-matching and merging two messy Indian-food datasets into one high-quality dataset — the data foundation behind the NutriVerse vision model.

Learning GitHub Through a Real Project

5 min read

Software Engineering /How a software-engineering course project became the moment Git and GitHub finally clicked — branches, pull requests, merge conflicts and collaborating like a real team.

[ 06 / CONTACT ]

LET'S BUILD
SOMETHING.

Location

Bangalore, Karnataka, India

Response

Usually within 24 hours

Open to

Internships · Research · Collaborations