aaditya@systems:~$curl -s https://aadilabs.fyi/whoami | jq .
Aaditya Joshi (AJ)
Software Engineer specializing in Distributed Systems, Query Optimization, and High-Throughput Data & AI Pipelines.
4+ years building and owning mission-critical production backends—from 0-to-1 international platform launches across Germany, Spain, and Belgium, to Dagster pipelines processing 100K+ vehicles, and AWS batch APIs collapsing 50 client calls into 1 with atomic idempotent retries.
Developers on AWS Builder Center platform
Connected vehicles across Germany, Spain & Belgium
75% drop via ORM profiling & materialized views
MS CS (Machine Learning & Scalable Systems)
Students mentored as Bayesian Statistics TA
Sustained high availability; <1 hr SEV1/2 MTTR
Interactive System Deep-Dives
Real-world distributed systems, query optimization breakthroughs, and streaming pipelines built and shipped to production. Inspect the architectural trade-offs, benchmarks, and interactive workflows below.
AWS Builder Center: 50-to-1 Batch API & Non-Account Invite Engine
Membership sync operations generated up to 50 independent REST calls per client action, causing network cascade latency, non-atomic partial failures, and high client-side retry orchestration. Furthermore, the platform had zero onboarding path for developers without an active AWS account.
AWS Builder Center is a critical builder engagement platform serving over 276,000 active users. Platform required membership synchronization and non-account acquisition under high-load enterprise environments.
Client dispatches 1 consolidated payload containing 50 user updates with idempotency token.
Payload: 50 items • 1 HTTP roundtrip
Lambda verifies tokens, executes DynamoDB TransactWriteItems ensuring all-or-nothing rollback.
Concurrency: 40 TPS resilience validated
1 response returned with per-record status. Zero partial state corruption; zero cascade retries.
Network reduction: 98%
Batch Processing Engine: Engineered batch REST API collapsing 50 single-record client roundtrips into a single idempotent payload with per-record atomicity and DynamoDB TransactWriteItems.
Shareable Invite Link Service: Built high-performance signed invite token service achieving ~200ms generation and ~150ms resolution latency, creating the platform's first non-account acquisition funnel.
Production Hardening & Compliance: Hardened with canary deployments, per-API CloudWatch alarms, automated GDPR (RTBF/DSAR) purge/export pipelines, and load tested to 40 TPS resilience.
Work Experience & Leadership
Leading engineering initiatives from early-stage system design through international multi-region production rollouts.
Software Developer Intern
Internship- Architected membership management batch REST API collapsing 50 single-record client calls into one payload with per-record atomicity and idempotent retries.
- Designed and implemented shareable invite-link service with ~200ms generation and ~150ms resolution latency, establishing the platform's first non-account acquisition path (100% production success in week one, 600+ generations, 200+ signups).
- Hardened production systems: canary deployments, per-API CloudWatch alarms, automated GDPR (RTBF/DSAR) purge/export pipelines, and load tested system resilience up to 40 TPS.
Lead Software Engineer — Fleet Intelligence
Full-time- Reduced core API endpoint latency by 75% (16s down to 4s) via targeted compound indexing, Django ORM query profiling, and materialized views for read-heavy workloads.
- Designed and deployed Dagster data pipelines processing 100s of GBs of hourly & daily telemetry from 100,000+ connected vehicles for predictive fleet intelligence.
- Integrated Claude and PydanticAI to automate metric visibility via a multi-step verification loop, slashing manual dashboard creation time by 60%.
- Maintained 99.999% platform uptime and resolved SEV1/SEV2 incidents in < 1 hour via automated CI/CD and production monitoring.
- Re-architected legacy backend services, modernizing ~200 outdated dependencies and resolving critical CVE security vulnerabilities.
- Led a team of 3 full-time software engineers and mentored 2 interns to complete feature ownership within 6 weeks.
Research & Development Intern
Internship- Engineered Arduino-based electric roll press prototype in C/C++ and developed PLC automation sequences for industrial packaging equipment.
- Benchmarked 10 distinct electrical components to optimize cost, electrical durability, and industrial operational reliability.
Academics, Teaching & Research
Rigorous foundations in Graduate Algorithms, Bayesian Inference, Machine Learning, and Distributed Scalable Systems.
Georgia Institute of Technology
- >Teaching Assistant for Bayesian Statistics mentoring ~400 graduate students per semester in probability, statistical inference, MCMC algorithms, and Bayesian hierarchical modeling.
- >Focus on distributed systems scalability, low-latency microservices, and modern ML/DL model optimization.
Indian Institute of Information Technology, Pune (IIIT Pune)
- >ACM-ICPC 2020 Nationwide Rank 935 / 3,200 top competitive programming teams.
- >Smart India Hackathon 2020 National Finalist for high-concurrency public service employment platform.
- >Active member of Algorithms & Systems Development cell.
1st Place Grand Winner for 'Someity' web accessibility browser engine.
Grand Prize Winner for best product demonstration and user accessibility impact.
National Finalist selected from tens of thousands of participants for public-sector challenge.
Nationwide competitive programming contest ranking 935 out of 3,200 teams.
Open Source Contributions & Products
Verified contributions to major open-source codebases, international hackathon-winning tooling, and high-concurrency platforms.
Someity
Engineered an open-source Chromium extension that transforms webpage accessibility in real-time without requiring source-code modifications. Implemented contrast remediation, text-to-speech, font reshaping, and screen-reading overlays.
Weblate Core Localization Engine
Contributed to the core codebase of Weblate (6k+ stars, web-based continuous localization system in Python/Django). Engineered a REST API endpoint to retrieve comment lists for translation units with filtering and pagination, improving programmatic CI/CD localization workflows.
High-Concurrency Job Portal (SIH Grand Finale)
Architected an end-to-end recruitment and candidate verification platform for the Smart India Hackathon 2020 Grand Finale. Implemented role-based access control, applicant tracking, and dynamic candidate screening under high concurrency.
Technical Stack & Systems Tooling
Languages, frameworks, data architectures, and infrastructure tested and delivered in high-load production environments.
- >Django & Django REST Framework (DRF)
- >FastAPI & Pydantic
- >Spring Boot & Hibernate
- >Celery & Distributed Task Queues
- >RabbitMQ & Kafka Event Streaming
- >RESTful API Architecture & Batch Endpoints
- >Microservices Concurrency & Idempotency
- >PostgreSQL (Indexing, EXPLAIN ANALYZE, Partitioning)
- >Amazon DynamoDB (Batch & Transactional Writes)
- >Redis (Hot Layer Caching, Session Stores)
- >Dagster (Orchestrated Asset Pipelines)
- >MySQL & MongoDB
- >AWS (Lambda, DynamoDB, CDK, CloudWatch, S3, IAM)
- >Docker & Containerization
- >Kubernetes & Pod Orchestration
- >CI/CD Pipelines (GitHub Actions)
- >Canary Deployments & Zero-Downtime Rollouts
- >GDPR Compliance (RTBF / DSAR Purge)
- >Grafana & Production Telemetry Monitoring
- >PydanticAI (Structured Agent Loops & Reflection)
- >Anthropic Claude & OpenAI APIs
- >MCP (Model Context Protocol)
- >PyTorch (Deep Learning Coursework)
- >Bayesian Inference & MCMC (Teaching Assistant)
- >RAG & Semantic Retrieval Architectures