Ashutosh Pandey
Backend Engineer building AI-powered systems and automation tools.
I build backend systems, AI-integrated applications, automation pipelines, and developer-focused products that solve real-world problems.
Who I am
I'm a Computer Science Engineering student at Indian Institute of Information Technology, Ranchi, focused on backend engineering and AI-integrated software development.
I enjoy understanding how systems work behind the scenes — from APIs and authentication to databases, automation pipelines, data processing, and LLM-powered applications.
I'm particularly interested in combining traditional backend engineering with modern AI capabilities to build practical software that solves repetitive and complex real-world problems — not demos, but systems people actually rely on.
Engineering Focus
Backend Engineering
FastAPI · REST APIs · Authentication · RBAC
AI Integration
LLMs · LangChain · LangGraph · RAG
Automation
Python Automation · Document Processing · Data Extraction · Validation Pipelines
Data & Databases
PostgreSQL · Supabase · MongoDB · Redis
Developer Tools
Git · GitHub · Docker · Postman
Academic background
Indian Institute of Information Technology, Ranchi
3rd Year8.00 CGPABachelor of Technology — Computer Science and Engineering
2024 – 2028
Coursework and self-directed projects centered on backend systems, databases, and AI-integrated software.
Class XII (Senior Secondary)
Class XII81%Higher Secondary Education
Completed
Class X (Secondary)
Class X86%Secondary Education
Completed
Current Engineering Focus
What I spend most of my time building with right now.
Backend Engineering
Designing REST APIs and service logic that stay maintainable as requirements change.
AI Integration
Wiring LLMs into real application logic instead of treating them as a black box.
Automation
Turning repetitive manual processes into reliable, reusable pipelines.
Data & Databases
Modeling data and querying it efficiently, sync or async.
Developer Tools
The everyday toolchain behind shipping backend work.
Technical skills
Languages
Backend
AI / GenAI
Databases
Data / ML
Web Scraping
Core CS
APIs & Integrations
Tools
How a request becomes a response
Backend concepts I understand and apply — hover a stage to see what happens there.
Browser / mobile app sends a request
Identity & Access
API Design
Data Layer
Shipping Software
Where I've worked
Engineering decisions and problems solved, not just a list of technologies.
AI/Automation Software Developer Intern · Ambition Colonisers Pvt Ltd
Onsite — Gurugram, Haryana
Built a multi-tenant financial reconciliation platform with schema-per-tenant data isolation, role-based access control, and an automated bank statement ingestion pipeline.
Problem
Reconciling financial data across multiple companies required strict data isolation between tenants and fine-grained permission boundaries, while bank statement handling was manual, format-inconsistent (PDF/XLSX/XLS/CSV), and didn't scale as more companies were onboarded.
Approach
Built a schema-per-tenant PostgreSQL architecture behind a FastAPI backend, layered RBAC with JWT authentication enforced at both the API and UI layers, and automated statement ingestion — including a Gmail-to-Drive collection step — with async background processing so large imports never timed out.
- Built a multi-tenant financial reconciliation platform (Python, FastAPI, PostgreSQL, React 18, Vite, Tailwind CSS) serving 8 isolated company schemas with schema-per-tenant data isolation
- Designed and implemented an RBAC system with JWT authentication and 4 permission tiers, enforced at both API and UI layers
- Built an automated bank statement ingestion pipeline supporting PDF, XLSX, XLS, and CSV formats, parsing 300+ statements with configurable page ranges and batch processing
- Engineered a Gmail-to-Google Drive automation using Google Apps Script and the Gmail/Drive APIs with OAuth 2.0, removing manual statement downloads
- Implemented asynchronous background job processing with real-time progress tracking and polling-based status updates to prevent timeouts on long-running imports
- Maintained 85+ automated integration tests validating business rules, permissions, and data integrity across tenant schemas
GenAI Intern · MySSCGuide
Remote
Built a data engineering pipeline that turns messy educational content from many sources into structured, validated JSON datasets.
Problem
SSC CGL/CHSL exam question data existed across 10+ scattered sources and formats — scanned PDFs, inconsistent layouts, bilingual (English–Hindi) content — with no structured, machine-readable form.
Approach
Combined web scraping, OCR-based document processing, and an ETL pipeline to extract, normalize, and validate the data, using an LLM with custom validation to assist with structuring question data at scale rather than relying on brittle manual rules alone.
- Engineered Python automation architectures that scaled content generation, reducing manual workload by 40% and saving 15+ hours weekly
- Built Python ETL pipelines ingesting data from 10+ web sources, processing 50K+ raw educational records into normalized formats
- Integrated LLMs with custom validation to generate structured educational content in JSON, reducing API schema errors by 30%
- Automated data extraction and AI-driven content workflows via OCR, improving data accuracy by 60% and generation speed by 40%
- Partnered with a 5-member team using Git/GitHub to resolve merge conflicts and integrate 10+ features
What I've built
Each one expands into the problem, the engineering decisions, and the tech behind it.
AI Doctor Assistant
Built during the IIIT Ranchi Quasar AI Hackathon by a 4-member team
LLM-backed symptom triage with a real API and database behind it.
TaskFlow
A full-stack task manager built to get backend fundamentals right.
Multi-Tenant Financial Reconciliation Platform
Built during the Ambition Colonisers Pvt Ltd internship
Schema-per-tenant automation platform for financial statement reconciliation across companies.
Educational Data Pipeline
Built during the MySSCGuide internship
Turning messy exam PDFs into clean, structured JSON.
Adaptive DSA Mock Test Platform
Currently Exploring / BuildingMock tests that adapt to what you're actually weak at.
Problems I've solved
Not tutorial projects — real constraints, real trade-offs.
I practice data structures, algorithms, and competitive problem-solving in C++ — 300+ problems solved across LeetCode, GeeksforGeeks, and CodeChef with a focus on Medium/Hard difficulty, plus 30+ CodeChef contests.
Problem
Financial reconciliation across multiple companies needed strict data isolation and role-based permission boundaries, with manual statement handling that didn't scale.
Approach
Schema-per-tenant PostgreSQL architecture + RBAC with JWT + an automated multi-format statement ingestion pipeline.
Outcome
A platform serving 8 isolated company schemas, parsing 300+ statements automatically and covered by 85+ integration tests validating permissions and data integrity.
System: Multi-Tenant Financial Reconciliation Platform
Problem
Large volumes of educational data scattered across 10+ inconsistent sources.
Approach
Web scraping + ETL pipeline + normalization + LLM-assisted validation.
Outcome
Structured, validated datasets (50K+ records) suitable for downstream applications, cutting manual workload by 40%.
System: Educational Data Pipeline
Problem
Processing scanned, bilingual educational PDFs into structured question data.
Approach
OCR + extraction + LLM processing + validation + duplicate filtering.
Outcome
Machine-readable, deduplicated question datasets, with data accuracy improved by 60% and generation speed by 40%.
System: Educational Data Pipeline