Shubham
Muthreja .
I'm an Engineering Manager and founder with 6+ years building production systems at scale. I managed a team of 6 engineers at Digital Paani, scaling a real-time IoT platform from 4 pilot plants to 55+ facilities processing millions of records daily. I'm currently building Qurkle — powered by MIRA, an LLM-based psychometric matching engine built in Python with FastAPI, LangChain, RAG, and vector embeddings.
The shortest possible summary of one engineer.
{
"name": "shubham",
"role": "Eng Manager · AI Systems",
"at": "@qurkle",
"from": "@India",
"open_to": [
"collaborations",
"research",
"coffee chats",
"nerdy discussions"
]
}Hello! I'm Shubham — an Engineering Manager and founder with 6+ years of experience building production systems at scale. I managed a team of 6 engineers at Digital Paani, scaling a real-time IoT platform from 4 pilot plants to 55+ facilities processing millions of records daily. My technical stack is TypeScript, Python, Node.js, React, AWS, and MongoDB — with recent deep work in LLM systems, FastAPI, LangChain, RAG pipelines, and vector embeddings. I'm open to EM, Tech Lead, and Staff Engineer roles at AI-first product companies.
I started my career building EV charging infrastructure at Numocity, moved into full-stack consulting, led development teams at BraynixAI, and eventually became Engineering Manager at Digital Paani — where I built 70%+ of the IoT platform from scratch before leading the team that scaled it to 55+ facilities serving clients like Delhi Jal Board, Amazon, Dr Reddy's, and Reliance. In February 2025, I co-founded Qurkle as CTO and have been building MIRA — a psychometric matching engine in Python and FastAPI that uses LLMs and vector profiles to understand people at a level far beyond what traditional apps do.
Things that shipped.
/* selected projects · sorted by recency · scroll → */
← scroll projects →
Qurkle
Consumer app powered by MIRA — an LLM-powered psychometric matching engine built in Python and FastAPI using LangChain, RAG, and 74-parameter vector profiles, matching users on deep personality signals rather than surface-level preferences. Shipped end to end: event creation, booking and ticketing flows, real-time notifications, and an analytics dashboard on AWS EC2 — with a Flutter app on the Play Store and a Node.js product API.
Digital Paani
Designed and built 70%+ of the IoT platform from scratch, then managed the team scaling it from 4 pilot plants to 55+ facilities for water treatment clients including Delhi Jal Board, Amazon, Dr Reddy's, Reliance, and BPTP — real-time dashboards, a SCADA-like visual plant control system built with React Flow, ML-based sludge settlement detection, and automated alerting, maintaining zero data loss.
Trusted Delivery
Digitised a 300+ truck logistics business end to end for a family enterprise — operations, HR, accounts, and a partner bidding marketplace handling requisitions for clients like PepsiCo and Blue Star, moved from paper to a MERN platform with Flutter apps and 12+ role-specific dashboards, preventing an estimated ₹16L/year in losses.
Emotional Chatbot
Chatbot integrated with real-time facial emotion detection using OpenCV and PyTorch. Detects user emotion from live camera feed and adapts responses dynamically for a more human-like interaction.
The path so far.
Building MIRA — an LLM-powered psychometric matching engine in Python/FastAPI using LangChain, RAG, and 74-parameter vector profiles to connect people on deep personality signals rather than surface preferences.
- ▸Built MIRA, an LLM-powered psychometric matching engine in Python and FastAPI using LangChain, RAG, and 74-parameter vector profiles, matching users on deep personality signals rather than surface-level preferences.
- ▸Shipped the complete product end to end: event creation, booking and ticketing flows, real-time notifications, and an analytics dashboard on AWS EC2 for reviewing 3000+ applications.
- ▸Set up CI/CD and Play Store deployment; led a Flutter engineer on UI while owning the Python/FastAPI inference service and Node.js product API end to end, including infrastructure and performance profiling, with typical API response times of 200-300ms.
- ▸Grew to 600+ accepted users, 200+ active on the production LLM pipeline, on a lean ₹1.15L marketing budget.
Managed a team of 6 engineers, scaling a real-time IoT platform from 4 pilot plants to 55+ facilities serving Delhi Jal Board, Amazon, Dr Reddy's, Reliance, and BPTP.
- ▸Managed a team of 6 engineers: hired, onboarded, and mentored across sprint planning, retrospectives, and performance reviews.
- ▸Scaled the platform from 4 pilot plants to 55+ facilities with up to 1,000 sensors per plant, serving Delhi Jal Board, Amazon, Dr Reddy's, Reliance, and BPTP; the system halved on-site plant manpower requirements.
- ▸Built a SCADA-like visual plant control system with the team using React Flow; drove code cleanup, latency reduction, and engineering standards across the codebase.
- ▸Built an ML sludge settlement sensor end to end: YOLOv8 on edge Raspberry Pi over an RTSP camera feed, trained on a 100K+ image dataset to a 0.95 F1 score, with custom fabricated hardware for automated sampling and flushing. Deployed across 10+ plants and later productised as a standalone offering.
- ▸Maintained zero data loss and roughly 5 hours total downtime per year across the fleet through instant rollbacks and AWS service alerting.
Took over the platform solo after the previous lead's exit; migrated the legacy broker stack to serverless AWS and built a real-time alert engine used as the demo product for the company's seed fundraise.
- ▸Took over the platform solo after the previous lead's exit; migrated the Mosquitto broker and Azure Tables stack to AWS Lambda ingestion, MongoDB time series, and EC2 with NGINX, eliminating per-plant broker servers and runaway storage costs.
- ▸Built a real-time alert engine evaluating time-sliced conditions on live sensor data with instant WhatsApp alerts, auto-escalation, and auto-resolution; plants moved from manual register rounds to 10-15 preemptive alerts a day handled by half the staff.
- ▸Designed a computed-sensor formula engine letting operators combine any sensors with arbitrary mathematical functions into virtual sensors, replacing manual Excel exports and synchronising readings from multiple Raspberry Pis per plant.
- ▸Cut latency by 30% on ingestion pipelines handling millions of records/day; wrote minute/hour/day time-series aggregations and revamped the dashboard UI.
- ▸The revamped platform was used as the demo product for the company's seed fundraise.
Led two cross-functional teams. Vision SDKs for edge devices. React/Next.js client work. AWS deployments on S3, CloudFront, Lambda. Reduced bug delays by 30%.
- ▸Collaborated with clients to revamp their websites by implementing modern, responsive designs and improving user experience using React.js/Next.js for the frontend and NodeJS with Express for backend functionality.
- ▸Conducted comprehensive SEO optimization for client websites by implementing best practices in meta-tagging, structured data, and performance enhancements to improve search engine rankings and boost website traffic.
- ▸Directed two cross-functional development teams within an Agile Scrum framework, managing daily stand-ups, sprint planning, and retrospectives to ensure team productivity and effective communication.
- ▸Reviewed the codebase extensively, focusing on React components, state management, and API integration to ensure high code quality and compliance with architectural standards.
- ▸Deployed projects on AWS cloud platforms using services like S3, CloudFront, and Lambda to deliver fast, secure, and cost-effective web applications.
- ▸Established and enforced robust development, testing, and release processes, reducing bug-related delays by 30% and significantly improving delivery timelines.
Digitised a 300+ truck logistics business end to end for a family enterprise, preventing ~₹16L/yr losses; built a partner bidding marketplace for clients like PepsiCo and Blue Star, MERN dashboards, and Flutter mobile apps.
- ▸Digitised a 300+ truck logistics business end to end for a family enterprise, moving operations, HR, accounts, and partner bidding from paper to a MERN platform with Flutter apps, preventing an estimated ₹16L/year in losses from missing truck parts and unnoticed bids.
- ▸Hired and led 3 developers; built a partner bidding marketplace handling requisitions for contracts with clients like PepsiCo and Blue Star.
- ▸Built truck part tracking and a last-mile delivery tracking app alongside 12+ role-specific dashboards covering operations, HR, accounts, and franchise management.
Built the flagship Android app in Java, then rebuilt it in Flutter with 4 whitelabeled versions shipped for enterprise EV charging clients. MQTT + LoRaWAN IoT integrations.
- ▸Built the flagship Android app in Java, then rebuilt it in Flutter; 4 whitelabeled versions shipped live for enterprise clients on the EV charging SaaS.
- ▸Built the battery swap flow (latch unlock, slot exchange) and implemented MQTT, LoRaWAN, and Bluetooth/Wi-Fi charger communication for weak-network environments.
- ▸Designed and implemented understandable architecture, increasing the code's maintainability and testability using the MVVM architecture pattern and writing test cases using JUnit and Mockito.
Graduated. Built a emotional chatbot as the final-year project — also wrote the paper for the same.
Let's build
something
useful.
// open to collaborations, research, and conversations about
// distributed systems / Psychology / LLM's / Comics / Physics / Nerdy stuff in general