EXP-01In build
No-code AutoML platform
AutoBench
Upload a dataset, train multiple models, compare leaderboards, and inspect explainability without writing ML code.
Full-stack ML platform Engineering AI systems into usable products
AI/ML Engineer · Backend Builder · Product-minded Developer
B.Tech CSE (IoT) student building deployable AI/ML products, real-time systems, and research-backed interfaces with production-grade polish.
Scroll
01
Engineering lives where curiosity meets execution.
I build the connective tissue between research, systems, and a product experience people can actually trust and use.
The throughline
“Build the model, understand the infrastructure, then make the interaction feel inevitable.”
2024
B.Tech CSE (IoT)
Started engineering foundation at IEM Kolkata with a focus on systems, data, and product execution.
2025
ML systems track
Built applied ML projects across fraud detection, churn prediction, and digit recognition.
2026
Research + product layer
Expanded into fog-computing RL research, backend APIs, and polished product experiences.
2027
Placement-ready engineer
Positioning for AI/ML, backend, and full-stack roles with production-grade portfolio artifacts.
0+shipped systems
0core disciplines
0product layers
02
Systems thinking, rendered as a graph.
Not a list of tools. A working constellation of the technologies I use to move from a question to a dependable outcome.
Selected capabilityPython
98% operating confidence
LanguageBackendFrontendAIInfraData
03
From experiments to experiences with teeth.
Six projects, each treated as a system: the problem, the interface, the architecture, and the proof that it works.
EXP-02Deployed
Risk intelligence dashboard
Credit Card Fraud Detection
Random Forest classifier on the ULB dataset with SHAP explainability and API-backed review flows.
AUC 0.9900EXP-03Research
Reinforcement learning research
HARD-RL Fog Load Balancer
Q-Learning load balancer for fog computing with fuzzy gating, adaptive rewards, and energy-aware epsilon decay.
-98.29% failed allocationsEXP-04Live
Vision model demo
MNIST Digit Recognizer
ANN and CNN digit classifier trained from scratch and deployed as a live interactive model demo.
98.93% accuracy04
Models are only useful inside a well-designed loop.
I think beyond the prompt or the metric: data quality, retrieval, evals, observability, and a product feedback cycle.
Explainable outputsMeasured evaluationHuman-in-the-loop
NEURAL PRODUCT PIPELINELIVE DESIGN MODEL
Stage 04Vector store
Models produce scored, inspectable outputs rather than magic.
05
Product confidence is an architecture decision.
Interfaces can be beautiful, but the system underneath must also be legible, resilient, and ready to evolve.
REFERENCE ARCHITECTURE system healthy
Current concernFastAPI services
Domain services and model operations
EXPERIMENT MEMORY tracking
06
Every model begins with an honest look at the data.
My workflow is designed to make useful signals visible early, validate decisions rigorously, and preserve why a result was trusted.
01EDA
02Feature design
03Training
04Validation
05Explainability
06Monitoring
07
Signals that make the work legible.
Proof matters when it is connected to how you work: shared code, focused practice, and a record of building across disciplines.
AN
01AI/ML Systems
Modeling, evaluation, explainability, and deployment practice.
AN
02Backend Engineering
FastAPI, Node.js, data contracts, and service architecture.
AN
03Product Interfaces
Interaction design, motion, accessibility, and performance.
Public repos9
Followers0
Stars (public)9
Primary languagePythonBuild focusAI + BackendRepo styleExperiments → Products
PROBLEM-SOLVING / CONTINUOUS
Deliberate practice is part of the build loop.
Applied problem-solving alongside systems work — patterns, data structures, state, and tradeoff awareness.
Problem-solvingDSA trackFocusPatternsModeConsistent practice
08
Have an ambitious system in mind? Let’s make it real.
I’m interested in AI/ML, backend, and product engineering work where craft and technical depth share the same brief.