Backend systems engineer specializing in distributed data platforms, cloud-native architectures, and production ML systems. I build scalable services that process millions of events, optimize infrastructure costs, and ship reliable software. Published researcher in AI security and collaborative machine learning.

class Engineer:
    def __init__(self):
        self.name = "Pratham Jain"
        self.role = "Backend Engineer | Data Platform Architect"
        self.location = "Gurugram, India"
        self.focus = ["Distributed Systems", "Data Engineering", "MLOps"]
        
    def current_work(self):
        return {
            "company": "Luminous Power Technologies (Schneider Electric)",
            "building": ["Real-time IoT data pipelines", "ML-powered analytics", "Cloud-native microservices"],
            "impact": ["14x query performance", "98% cost reduction", "100K+ users scaled"]
        }

AWS: S3 • EC2 • Lambda • Bedrock • DynamoDB
GCP: BigQuery • Dataflow • Vertex AI
Azure: Data Factory • Databricks • Event Hub • AKS • DPS • Notification Hub

Luminous Power Technologies (Schneider Electric Group) | Feb 2025 – Present

Infrastructure Optimization:

• Migration: Java 8 cron → Azure Data Factory + Azure Batch ETL
• Performance: 14× faster queries, 98% cost reduction
• Architecture: 12-hour micro-batches, repartitioning, point reads

Scalable Data Pipelines:

• Built: Deduplication & validation pipelines with tracking IDs
• Scaled: 100K users (~5.1M notifications/month) via Azure Notification Hub
• Implemented: Idempotent event handling for Event Hub

Distributed Systems:

• Problem: Thundering herd on Azure DPS
• Solution: Exponential backoff with jitter, idempotent retries
• Result: 11.3 hours → 32 minutes provisioning, 99.8% fewer retries, 100% success

Machine Learning:

• EL image segmentation: 24 defect classes, 67% avg IoU (AMP + distributed training)
• Physics-informed network: KPI extraction from IV curves, 95% accuracy
• Quality grader: Clustering-driven approach

Cloud Services:

• Deployed: Adaptive-agent & LLM-backed services on AKS
• Features: Redis caching, idempotent workflows
• Impact: 78% lower latency, 88% faster responses, 30% cost savings, 7% higher satisfaction (10K+ users)

Neocfo.io | Feb 2025 – Apr 2025

• Built multi-agent backend using LangGraph and Amazon Lex for natural-language legal analysis
• Handled 2,000+ daily requests for legal clients
• Migrated AWS Lambda → EC2, achieving 40% cost reduction while maintaining availability
• Implemented ML pipelines for legal document analysis and natural-language querying

Bosch Global Software Technologies | Mar 2024 – Feb 2025

• Trained quantized & pruned TinyML models on 300+ GB IoT sensor data (HVAC systems)
• Delivered 15.5% energy savings with statistical significance
• Built containerized distributed backend for multimodal diagnostic platform
• Achieved 96.8% diagnostic accuracy on 500+ FNAC images
• Engineered fault-tolerant systems for STM32 microcontrollers with real-time guarantees
• Deployed services on AWS (S3, Lambda, EC2) with containerized workloads and autoscaling

Open-source npm and PyPI package for storing and sharing non-code artifacts outside version control. Replaces ad-hoc sharing across chat tools and personal drives.

Key Features:

• Simple, auditable storage layer with stable identifiers
• Metadata support and idempotent operations
• Streaming-safe APIs
• Integrates with cloud object storage and existing IAM

Flipkart Grid 6.0 - Level 2 Finalist

Hybrid edge-cloud system for automated product detection using PyTorch and Qwen-VL2-2B.

Achievements:

• >95% detection accuracy
• <1s inference latency
• Docker-based microservices architecture

Luminous TechnoX - First Runner-Up

IoT-driven energy optimization platform with LSTM-based tariff prediction and MILP-based scheduling.

Results:

• 30% energy savings
• Automated pipelines and monitoring
• Built on AWS IoT, Lambda, S3

2025 | Multimodal imaging and FNAC

Multimodal system combining mammography and FNAC to improve diagnostic yield.

2025 | MLOps & Edge Computing

Pipeline design for scalable model training, edge packaging, and automated deployment.

2025 | AI Security

Architected a tamper-resistant protocol for collaborative model updates using blockchain primitives.