Harsh’s DevSpace — /brain/src/hero.js
hero.jsabout.jstimeline.jsprojectsexpertise.jscontact.jsresume.pdf

Welcome

I’m Harsh N. Chandak,
• M.S. Computer Science @ ASU (4.0)
• Full-stack engineer: product & AI systems
• Ships fast, scales clean & breaks nothing (mostly)

resume.pdf ↓LinkedIn

(My name means "Happiness" in Hindi. Yes, I'm aware of the irony.)

"Good code is like a good joke: if you have to explain it, you messed up."
– Harsh N. Chandak
1
// About Me
2
const dev = {
3
name: "Harsh N. Chandak",
4
title: "Full-stack Developer | AI Engineer | Research Assistant",
5
education: "MSCS @ASU",
6
philosophy: "Design for failure, measure the boring parts, and automate what should not need heroics",
7
core_skills: [
8
"Full-stack Development",
9
"AI Systems & LLM Pipelines",
10
"System Optimization",
11
"System Design",
12
"Telemetry & Analytics"
13
],
14
traits: [
15
"system-thinker",
16
"api-gardener",
17
"legacy-fixer"
18
],
19
daily_setup: [
20
"☕ Coffee",
21
"Postman",
22
"VS Code",
23
"GPT & Claude",
24
"too many browser tabs"
25
],
26
tech_stack: [
27
"JavaScript",
28
"TypeScript",
29
"Python",
30
"React",
31
"Next.js",
32
"Node.js",
33
"Express.js",
34
"FastAPI",
35
"PostgreSQL",
36
"MongoDB",
37
"Redis",
38
"Docker",
39
"Kubernetes",
40
"AWS",
41
"Kafka",
42
"Git",
43
"MySQL"
44
]
45
};
> git log --oneline ▍
d129bea Software Engineer — AI Systems & Product @ ASU
Author: Dev Harsh <harsh-chandak>
Date: Dec 2025 – Present
Turns out making AI behave consistently at scale is a completely different problem than making it work once.
c3dd2f2 Software Engineer — Developer Tooling @ ASU
Author: Dev Harsh <harsh-chandak>
Date: Aug – Dec 2025
Nobody filed a ticket for the observability gap. I just noticed it, scoped it, and built the whole thing.
79b5dd7 M.S. in Computer Science @ Arizona State University
Author: Dev Harsh <harsh-chandak>
Date: Aug 2024 – May 2026
6ea7bd3 Software Engineer — Full-Stack & Product @ Neuromonk
Author: Dev Harsh <harsh-chandak>
Date: Jan 2023 – Jun 2024
Early-stage startup, real production load, 70+ clients depending on the system. No safety net.
b61862f Freelance Web Developer
Author: Dev Harsh <harsh-chandak>
Date: Jul 2022 – Dec 2023
Working directly with founders taught me that vague requirements are just undiscovered scope creep.
cd96a2c Intern @ ISOBEX LLP
Author: Dev Harsh <harsh-chandak>
Date: Jul 2021 – Aug 2021
First time touching production code. Refactored queries nobody wanted to touch and cut latency in half.
1a3fe94 Bachelor of Technology @ Pune University
Author: Dev Harsh <harsh-chandak>
Date: Jul 2019 – May 2023
PS C:\Users\Harsh\brain\src\projects> ▍

"Multi-Agent AI Pipeline with Evaluation Infrastructure"

Featured

// Production system built at Arizona State University.

const period = "Dec 2025 – Present";
const stack = [ "Python", "FastAPI", "WhisperX", "OpenAI API", "Gemini API", "Docker", "AWS S3" ];

Processed 10K+ pipeline runs through diarization, LLM refinement, and multi-stage quality evaluation with reproducible artifacts.

Visible Impact

  • 10K+ evaluation runs used to surface model failure modes.
  • Improved output consistency by about 20% with scoring thresholds and retry logic.
  • Containerized FastAPI services with S3 artifact storage and run logging.

How It Worked

  • Integrated WhisperX diarization, prompt refinement, and multi-candidate image generation end to end.
  • Designed rubric-based scoring with deterministic selection and quality thresholds.
  • Built a workflow that made experiments easier to debug instead of just easier to run.
Codebase is private because the system was built in a university production environment.

"Developer Telemetry & Observability System"

Featured

// Internal developer tooling built at Arizona State University.

const period = "Aug – Dec 2025";
const stack = [ "TypeScript", "Express", "React", "Vite", "PostgreSQL", "AWS" ];

Built an end-to-end telemetry platform covering event schema, ingestion, storage, and dashboarding for internal developer analytics.

Visible Impact

  • Owned the system from schema design through dashboard UX.
  • Handled 2K+ events/day with bearer auth, validation, and idempotent state tracking.
  • Reduced P95 latency by 45% via async batching and backpressure.

How It Worked

  • Closed an observability gap that teams previously handled with scattered logs and manual checks.
  • Designed a storage model that made experiment tracking and performance monitoring queryable.
  • Shipped a React dashboard so data was visible to non-backend users too.
Private internal tooling; public screenshots are not available yet.

"Job Alerts + Application Tracker"

Featured

// A personal project born from job hunt frustration.

const period = "Summer 2025";
const stack = [ "Next.js", "MongoDB", "Puppeteer", "Discord Webhook", "JWT Auth", "Tailwind CSS" ];

Automated job discovery, alerting, and application tracking so new openings reached me in real time instead of after manual search loops.

Visible Impact

  • Scraped dynamic job boards with scheduled Puppeteer jobs.
  • Filtered roles by remote, stack, and title before sending Discord alerts.
  • Added JWT auth and multi-user application tracking workflows.

How It Worked

  • Normalized scraped data into structured records for filtering and search.
  • Cut repetitive manual job-search time into a single alert-driven workflow.
  • Shipped secure auth and scoped data storage instead of a single-user script.
[Live Demo]Live demo available.

"Kafka + Neo4j Streaming Pipeline"

// ASU course project on distributed systems.

const period = "Spring 2025";
const stack = [ "Kafka", "Neo4j", "Kubernetes", "Docker", "Python", "Spark" ];

Built a Kafka to Neo4j streaming pipeline for live graph analytics, influence scoring, and event-driven updates under load.

Visible Impact

  • Handled 5K+ events/min with sub-second updates on Minikube.
  • Processed graph analytics across 1.5K+ nodes with BFS and PageRank.
  • Improved throughput about 30% by tuning batch sizes and concurrency.
[Proof Artifact]Architecture animation included below.

"Spatial Data Analysis using Apache Spark and Scala"

// ASU big data and geospatial analytics course.

const period = "Spring 2025";
const stack = [ "Apache Spark", "Scala", "Java", "Spatial SQL Queries" ];

Processed large geospatial datasets with Spark and Spatial SQL to speed up hotspot detection, joins, and spatial filtering.

Visible Impact

  • Processed 1M+ geospatial points with custom RDD pipelines.
  • Reduced join runtime by about 40% using geohash-based aggregation.
  • Built visual output for hotspot and traffic-pattern analysis.
[Proof Artifact]Visualization snapshot included below.

"Mapping Accident Trends & Patterns"

// ASU D3.js data storytelling project.

const period = "Fall 2024";
const stack = [ "D3.js", "JavaScript", "Node.js", "GeoJSON" ];

Mapped 185K+ accident records into interactive geographic and temporal views, finishing in the top three of the course showcase.

Visible Impact

  • Parsed and grouped 185K+ records for spatial analysis.
  • Built six interactive views including heatmaps, timelines, and radial charts.
  • Placed 3rd in class for storytelling, usability, and clarity.
[Live Demo]Live demo available.

"Warehouse Robot Optimization using Clingo"

// ASU Knowledge Representation and Reasoning project.

const period = "Spring 2025";
const stack = [ "Clingo", "Answer Set Programming", "Python" ];

Modeled warehouse routing and delivery constraints in ASP, then wrapped the solver in Python for dynamic horizon planning.

Visible Impact

  • Encoded multi-robot planning over a constrained warehouse grid.
  • Minimized makespan in about 80% of test cases.
  • Ran about 2x faster than a brute-force Python baseline.
[Live Demo]Live demo available.
// cat expertise.js

Target Roles

Backend / Platform EngineeringAI Systems / LLM InfrastructureProduct-minded Software Engineering

Best fit for teams hiring around backend systems, data-heavy services, platform tooling, and production AI workflows where scale, failure handling, and observability matter.

backend.ts

Backend & APIs

I build APIs and services that are observable, resilient, and easy to extend.

Node.jsExpressFastAPITypeScriptPostgreSQLMongoDBRedis
ai-systems.ts

AI Systems

Strongest work: evaluation loops, retry systems, model orchestration, and artifact tracking.

PythonOpenAI APIGemini APIWhisperXPrompt evaluationRun logging
infra.ts

Data & Infrastructure

Comfortable with event pipelines, containers, distributed processing, and cloud deployment.

KafkaNeo4jDockerKubernetesAWSSparkTelemetry
product.tsx

Frontend & Product Delivery

I can ship the interface too, but I position myself around the systems underneath it.

ReactNext.jsViteTailwind CSSDashboard UXAuthentication
//contact.js▍
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
const contactInfo = {
}
function sendMessage() {
}

© 2025 Harsh Nitinkumar Chandak. All rights reserved.

Email: harshnchandak@gmail.com

Harsh@dev • mainLn 42, Col 7 • UTF-8 • LF