Soujanya GuptaSOFTWARE ENGINEER · AI AGENT EVALUATION

INDORE, INDIAIST

FRONTEND PLATFORMS & AI AGENT EVALUATION

BUILD.TEST.Understand.

I build frontend products and AI evaluation systems at Ethara AI—from MyMandi and Kuberha dashboards to model-testing tools.

Explore my work View résumé ↗

EXPLORE THE BUILDSMorrow ↗MyMandi & Kuberha ↗100+ business-app integrations across both products

01 / SELECTED WORK

IDEAS, TAKEN ALL THE WAY.

PROOF OF
the work.

Production frontend platforms, AI evaluation infrastructure, and selected personal projects.

PRODUCTS / EVALUATION / ENGINEERING

FEATURED PERSONAL BUILD / 01 LOCAL WORKSPACE · FREE BETA

morrow

A little direction.A better tomorrow.

A career workspace that puts learning first. Study, shape your résumé, manage applications, and follow your AI agents—all from your own computer.

Product & frontend
Agent orchestration
Python · React · SQLite

Explore the code ↗

A LOOK INSIDE THE PRODUCTCHOOSE A VIEW BELOW ↓

morrow / LaunchpadLOCAL BETA

YOUR NEXT CHAPTERSample workspace

Launchpad

Make your next move.

YOUR CAREER, IN VIEWA little progress.
Every single day.Applications & next steps ↗

Your daily flow1 Learn & research2 Prepare your profile3 Find matching roles4 Review the report

Activity overviewIllustrative sample

MONTUEWEDTHUFRISATSUN

Applications, daily flows, and the record of what happened.

02 / IN PRODUCTIONETHARA AI

BUILT FOR
the everyday.

10MONTHS

Two products.
One engineering team.

My frontend work across dashboards,
orders, integrations, and AI.

01 / COMMERCE & FULFILLMENT

MyMandi

FRONTEND ENGINEERING

Local commerce, connected.

DASHBOARD → CONNECTED WORKFLOWSUI ILLUSTRATION

I built and maintained the sales, inventory, and order-management dashboard, including Shiprocket delivery workflows and AI features.

SalesInventoryOrdersDelivery

02 / AI & RETAIL OPERATIONS

Kuberha

FRONTEND ENGINEERING

Retail operations, made usable.

DASHBOARD → CONNECTED WORKFLOWSUI ILLUSTRATION

I built the website and admin dashboard for sales, inventory, and orders, with role-based access, Shiprocket delivery, and AI workflows.

AdminStockOrdersApps & AI

100+

THE CONNECTED SIDE OF THE WORK

Less switching.
More connected.

Business-app integrations across both products.

Google SheetsConnected data
Email appsIllustrated with Gmail
Orders & delivery
Business APIsConnected workflows
AI featuresIn the working interface
Inventory & ordersAcross the dashboard

PROFESSIONAL WORK / ETHARA AI

AI AGENT EVALUATION · INTERNAL PLATFORM

WildClawBench fork · OpenClaw harness

Test behavior.
Not just answers.

I expanded an open-source benchmark into an internal platform for evaluating frontier models, with multi-provider routing, state-drift tests, evidence-based grading, and complete cost tracking.

View résumé ↗

MODEL REQUEST → AUDITABLE RESULT

Claude on AWS Bedrock · GPT-5.5 · OpenRouter

ROUTINGLiteLLM sidecarPrompt caching + per-request cost

EXECUTIONAgent runs101 mock services
Injected state drift

GRADINGThree-judge councilEvidence budgets
Weighted pytest suites

Trajectories, scores & costBounded orchestration · Docker isolation · EC2 delivery

ARCHITECTURE ILLUSTRATION / INTERNAL IMPLEMENTATION

101: FastAPI mock services
3 judges: Evidence-based council grading
5 agents: Maximum concurrent sub-agents
Per request: Cost tracking and auditability

001 / THE INTELLIGENCE PROJECTPYTHON · LANGCHAIN · MULTI-AGENT SYSTEMS

EvalAgentFlow↗

FROM NOISE.
TO know-how.

An eight-agent pipeline turning CSV evaluation data into benchmark reports, combining human rubrics with CodeBLEU, n-gram similarity, and dataflow matching.

01 — ORCHESTRATE02 — EVALUATE03 — UNDERSTAND

READY WHEN YOU ARE↗

A little order from the chaos.

INTERACTIVE PIPELINE ILLUSTRATION

INTELLIGENT SYSTEMS, HUMAN-READABLE RESULTS.EXPLORE THE BUILD ↑

002 / THE LEARNING PROJECTTYPESCRIPT · NEXT.JS · VERCEL

cohart

YOUR NEXT
BIG chapter.

An episode-based study platform for React, Next.js, backend engineering, generative AI, DSA, and system design. Built with TypeScript and Next.js.

LEARN AT YOUR PACE.BUILD ON WHAT YOU KNOW.

YOUR NEXT CHAPTER↗

cohart

Big things start
with small steps.

THE WEB DEVELOPMENT PATH

Build the whole stack.

01React & Next.js architecture↗
02Node.js & MongoDB backends↗
03System design fundamentals↗

INTERACTIVE CURRICULUM CONCEPT

A LITTLE STRUCTURE. A WORLD OF POSSIBILITIES.EXPLORE COHART ↗

003 / THE REAL-TIME PROJECTMERN · WEBSOCKETS · REDIS

RideMate

NEARBY RIDES.
Live connections.

Geospatial ride matching with live updates, Redis caching, and a dedicated women-to-women matching mode.

MONGODB / GEOSPATIAL MATCHINGNODE.JS / CONCURRENT PROCESSING

RIDERNEARBY MATCH

LocationMongoDBLive match

WEBSOCKETS · REDIS · WOMEN-TO-WOMEN MODE

ARCHITECTURE ILLUSTRATION / NOT A LIVE RIDE SERVICE

LOCATION → MATCHING → LIVE UPDATESVIEW SOURCE ↗

AND A FEW MORE
THINGS IN PROGRESS.Follow the rabbit hole on GitHub.

02 / EXPERIENCE, IN THE MAKING

EVERY CHAPTER LEAVES SOMETHING WITH YOU.

BUILT BYevery chapter.

Software Engineer at Ethara AI
(formerly Green Rider Technology).
Mar 2024–present · Gurugram, India.

THREE AREAS OF OWNERSHIP. ONE CURRENT ROLE.

THE THREAD I CARRY FORWARD

Test behavior when tools
and backend state change.

THE WORK. THE RESPONSIBILITY.SELECT A CHAPTER ↓

ETHARA AI / MAR 2024–PRESENT

Agent evaluation platform

Software Engineer · WildClawBench / OpenClaw

Expanded a WildClawBench fork into an internal, multi-provider agent-evaluation platform.

Routed Claude on AWS Bedrock, GPT-5.5, and OpenRouter through LiteLLM with prompt caching and per-request costs.
Built three-judge council grading and weighted pytest suites, plus drift injection across 101 FastAPI mock services.
Added bounded multi-agent runs, audit trails, Docker sandboxing, and EC2 delivery.

PythonFastAPILiteLLMAWS Bedrock

ETHARA AI / SAME SOFTWARE ENGINEERING ROLE

Model & dataset quality

Model evaluation · Training-data quality

Wrote domain prompts and evaluation rubrics for a Meta LLaMA 4 Maverick fine-tuning program.

Built Python and LangChain pipelines benchmarking OpenAI, Gemini, Claude, LLaMA, and Grok across finance, commerce, and healthcare.
Owned taxonomy classification, entropy checks, inter-rater agreement, and NLP preprocessing.
Built an internal Odoo app for labeling golden datasets.

LangChainEvaluation rubricsDataset quality

ETHARA AI / 10-MONTH PRODUCT ENGAGEMENT

Frontend platforms

MyMandi · Kuberha · Model-training UI

Built and maintained the sales, inventory, admin, and order-management dashboards for MyMandi and Kuberha as part of the Ethara AI engineering team.

Delivered frontend workflows for 100+ business-app integrations across both products, including email applications and Google Sheets.
Connected Shiprocket order and delivery workflows and AI features in both platforms.
Built the Kuberha website and role-based admin dashboard in Next.js, Tailwind CSS, and shadcn/ui.
Built a model-training UI with reusable React components, real-time metrics, and code splitting.

Next.jsReactShiprocketApp integrations

FEB 2021–APR 2021 / INTERNSHIP

Robotronix Engineering Tech

Full-Stack Developer Intern

The sole developer on Indore Carbonic, a carbon-product ordering platform.

Built and shipped the product end to end with React.js and REST APIs, taking responsibility for the interface and ordering workflow.

React.jsREST APIsEnd-to-end delivery

03 / PUBLIC CODE

OPEN THE REPO. FOLLOW THE WORK.

BUILT IN
the open.

Products, experiments, and the commits behind them. Explore my work where it’s built.

Visit my GitHub profile ↗

THE PERSON BEHIND THE COMMITS

Soujanya Gupta

@GuptaSoujanya ↗

GITHUB ACHIEVEMENTS04 / EARNED

A few milestones along the way.
The work keeps going.

Explore achievements

github.com / GuptaSoujanya

FROM MY PUBLIC PROFILE CAPTURED 28 OCT 2026

SELECTED REPOSITORIESREAD THE CODE. SEE THE PROGRESS.

Morrow ↗

PERSONAL PROJECT / BETA

A local career workspace for study plans, résumé tailoring, browser-assisted workflows, and activity tracking.

Python · JavaScript · Local workspaceMy commits ↗

llm-eval-harness ↗

EVALUATION PROTOTYPE

Task interfaces and evaluation metrics for LLM experiments. The current CLI uses a mock model provider.

Python · Task suites · Metric interfacesMy commits ↗

agentic-rag-benchmark ↗

RETRIEVAL PROTOTYPE

A small in-memory retrieval and evaluation prototype with recall@k and NDCG metrics.

Python · Retrieval · EvaluationMy commits ↗

inference-latency-profiler ↗

TIMING SIMULATION

A simulated token-stream profiler for time to first token, inter-token latency, and throughput.

Python · Streaming metrics · Simulated delaysMy commits ↗

04 / THE HUMAN

A LITTLE CONTEXT BEHIND THE CODE.

PART ENGINEER. PART “WHAT IF?”

Wired to build.
Made to wonder.

I’m Soujanya, a software engineer based in Indore. I build frontend platforms and the evaluation systems that help teams understand AI agents.

At Ethara AI, I work across MyMandi and Kuberha, model evaluation, training-data quality, and the web interfaces around those systems. I care about rigorous engineering and interfaces people can use with confidence.

A little more about me

THE PERSONAL LAB OFMEMBER / 001

SOUJANYA
GUPTASOFTWARE ENGINEER
FRONTEND & AI SYSTEMS

INDORE, INDIA ↗

A WORK IN PROGRESS. ALWAYS.

01 / HOW I THINK

Untangle the complex.

Clarity first. The code follows.

02 / HOW I BUILD

Care for the details.

The small things make the whole thing.

03 / HOW I GROW

Leave room for play.

A good “what if?” is usually worth following.

THE EVERYDAY TOOLKIT

Across the interface and the system.

Languages: Python · JavaScript · TypeScript · C++ · SQL
AI & evaluation: OpenClaw · WildClawBench · LangChain · LLM-as-judge councils · Multi-agent orchestration · RAG · Prompt engineering · CodeBLEU · Dataset quality · NLP
Web engineering: React · Next.js · Node.js · Express · FastAPI · Tailwind CSS · shadcn/ui · Redux Toolkit · TanStack Query · WebSockets
Data & infrastructure: AWS Bedrock · LiteLLM · Docker · EC2 · MongoDB · MySQL · Redis · SQLite · pytest · REST APIs · CI/CD · Git LFS

EDUCATION / 2020–2024

B.Tech, Computer Science & Engineering

Medi-Caps University

CERTIFICATIONS

AWS Academy Cloud Architecting
AWS Academy Cloud Foundations

05 / THE NEXT GOOD THING

STARTS WITH A CONVERSATION.

HAVE A GOOD
what if?

A NOTE TO SOUJANYALet’s make
something matter.DROP ME A LINEA PROJECT. AN OPPORTUNITY.
OR THE BEGINNING OF SOMETHING GOOD.

THE INBOX IS OPEN

semrisoujanya2003@gmail.com

Soujanya GuptaSOFTWARE ENGINEER · AI AGENT EVALUATION