Profile
Menu
Switch Theme

© 2026 Atharva Deshpande

Know Me Better

About Me

About Pic

I'm a Software Engineer moving into P2P analytics, using process mining to turn process data into improvements business teams can act on!

I spent 38 months at Bentley Systems across 4 product teams, delivering 25+ features while holding 85% test coverage and no user-reported production bugs. I'm now pursuing an MBA (PGPM) at GLIM Chennai, where I analyze Procure-to-Pay and lead-to-order event logs, configure ERPs like Odoo and SAP S/4HANA, and apply machine learning and LLM-assisted workflows to real business problems. I bring engineering discipline and business-process thinking together to turn data into decisions.
Name:

Atharva Deshpande

Location:

Pune, Maharashtra, India

Email:

atharvadeshpande174@gmail.com

Availability:

Open to Work

Explore My Work

Projects

Process Mining: P2P, Reimbursement and Lead-to-Order view 1Process Mining: P2P, Reimbursement and Lead-to-Order view 2Process Mining: P2P, Reimbursement and Lead-to-Order view 3
01

Process Mining: P2P, Reimbursement and Lead-to-Order

Process MiningPM4PyPower BIPythonP2P
  • Analyzed a P2P event log and found invoice clearing took a median 41 days (38% of waiting), so proposed a pilot clearing run
  • Found 28% of cases payment-blocked (18% of waiting), so recommended rule-based release on matched receipts
  • Mapped reimbursement and lead-to-order flows from event logs, built a multi-agent PM4Py workflow and a Power BI dashboard
ERP Capstone: Process Mapping and LLM-Assisted Config view 1ERP Capstone: Process Mapping and LLM-Assisted Config view 2
02

ERP Capstone: Process Mapping and LLM-Assisted Config

OdooBPMNLLM AutomationERP
  • Configured Odoo for a grocery firm (two warehouses, FEFO, lot tracking) through LLM-guided automation
  • Delivered BPMN models, solution design, and a go-live readiness report
  • Generalized the approach into a reusable LLM framework
Data Mining and Machine Learning view 1Data Mining and Machine Learning view 2
03

Data Mining and Machine Learning

PythonMachine LearningRandom ForestNeural Networks
  • Designed a multi-agent LLM workflow (profiling, cost matrix, model runs, review, reporting) to evaluate any dataset using python
  • Scored models on the real cost of wrong decisions using a cost matrix
  • Applied it to 30,000 credit card records, comparing Random Forest and Neural Network for missed payments, with human-in-loop verification.
Website Builder view 1
04

Website Builder

Multi-Agent AIClaude CodeGoogle Antigravity
  • A Product Manager agent interviews the client and writes site.spec.json; specialist agents then build and maintain the site.
  • Guardrails are enforced by deterministic scripts, not prompts.
  • Built and tested on Claude Code and Antigravity, and generalised from the Kirana app.
Process Mining Agent Workflow view 1
05

Process Mining Agent Workflow

PM4PyPythonGoogle AntigravityProcess Mining
  • Multi-agent process-mining workflow for Google Antigravity, backed by a Python toolkit built on PM4Py.
  • Produces its full process-mining report in both Markdown and PDF.
  • Tested on a Kaggle incident event log, then run on the public BPI Challenge 2019 Procure-to-Pay log.
FinConsent view 1
06

FinConsent

FinTechPrototype
  • Trade-authorization and tax-impact platform for SEBI-registered investment advisers.
  • Proof of concept delivered as a website or app together with a client deck.
  • Also the subject of my AI PM Portfolio Website assignment (Milestones 1 and 2).
Role Craft view 1
07

Role Craft

Multi-Agent AIResume Tailoring
  • Multi-agent AI platform that tailors your resume to the specific job posting you are applying to.
  • Also manages the job-application pipeline, alongside the resume tailoring.
  • Built solo as a personal and academic project, from first idea to a working platform.
Task Planner view 1
08

Task Planner

Product DesignTask Management
  • Built for group work after deadlines kept getting scattered across WhatsApp, drives and notebooks.
  • Brings priority tags, deadline reminders and a shared progress view into one place.
  • Designed, built and watched in use by classmates; the features they ignored were cut.
Credit Card Default Prediction view 1Credit Card Default Prediction view 2
09

Credit Card Default Prediction

PythonRandom ForestNeural NetworkLogistic Regression
  • Compared Random Forest, Neural Network and Logistic Regression on 30,000 Taiwanese credit card customers, scored on expected cost.
  • Random Forest caught 72% of customers who miss payments (37% hit rate, ROC-AUC 0.768) but was flagged for overfitting.
  • Expected cost: Random Forest NTD 5.25M, Logistic Regression NTD 5.64M (accepted), Neural Network NTD 6.0M (rejected by reviewer).
PodFast view 1PodFast view 2PodFast view 3PodFast view 4PodFast view 5
11

Dee Foundation

ReactViteFirebaseTailwind CSS
  • Official platform for a non-profit organization dedicated to community development across India.
  • Showcases initiatives in STEM education, healthcare, vocational training and women empowerment.
  • Serves as the digital hub for their 7-year journey and community-driven development (CDD) programs.

My Resume

Resume

Atharva Deshpande

Software Engineer

atharvadeshpande174@gmail.com +91 9890029809 Pune, Maharashtra, India GitHub LinkedIn
38Months of experience
25+Features delivered
85%Test coverage held
4Product teams
Summary

Software Engineer with 38 months of experience across 4 product teams at Bentley Systems, currently pursuing an MBA (PGPM) at GLIM Chennai. Looking to build a career in P2P analytics transformation, using process mining to turn process data into improvements business teams can act on.

Experience

Software Engineer 1

Bentley Systems
August 2022 - October 2025Pune, Maharashtra, India
  • Delivered 25+ features across 4 product teams and earned an early promotion for consistent delivery and ownership.
  • Ran Scrum sprints with requirements clarified up front and demos every sprint, holding 85% test coverage and no user-reported production bugs.
  • Delivered an assets data-synchronization feature end to end, integrating a new public API and writing its integration tests.
  • Owned an internal data-schema explorer (WIMSE) end to end as sole developer for 3 months, clearing security-audit and scorecard failures.
  • Led development of a project creation workflow as Lead Developer over 6 months (2025).
Education

MBA (PGPM)

2026 - 2027
Great Lakes Institute of Management, Chennai | PGPM Batch of 2026 - 2027
CGPA: 3.02
  • Team Lead: Campus dining analytics, Empirical Study Project (2026-27, ongoing)
  • Part of Product Management Club
  • Participated in Ideathon hackathon organized by GLIM Chennai

Bachelors in Computer Engineering

2018 - 2022
PVG's College of Engineering and Technology, Pune | Savitribai Phule Pune University
Percentage: 85.52%
  • Web Team Lead, Association of Computer Engineering Students (ACES PVGCOET), 2 years
  • Committee Work: Developed one of the virtual labs for V-Labs, IIT Bombay
Skills
Process & ERPP2P / O2C Process KnowledgeERP: Odoo, SAP S/4HANAProcess Mining and BPMNWorkflow Automation
Data & AnalyticsPythonMachine LearningStatistics and Financial AnalysisExcelTableauPower BISQLMongoDBData Storytelling
FrontendReactNext.jsTypeScriptJavaScript (ES6+)ReduxReact QueryHTML5CSSMFEsModule Federation
TestingJestReact Testing LibraryCypressPuppeteer
LanguagesEnglish (Advanced)Hindi (Advanced)Marathi (Advanced)
Certifications
  • SAP S/4HANA Hands-On Training (2026)
  • Neural Networks and Deep Learning, DeepLearning.AI (2020)
  • Python Training, Spoken Tutorial, IIT Bombay (2021)
Publications
  • Springer: Extractive Audio Summarization (2022)
Activities
  • Employee Stock Ownership Plans Award, Bentley Systems (2023)
  • Winner, National Level Hackathon (2020)
  • Volunteer: rebuilt the Dee Foundation website layouts