Skip to main content
Aashish Kumar

Aashish Kumar

AI/ML Engineer & Technology Lead

Talent
PuneMember since September 20, 2026
PythonSQLNumPySeabornMatplotlibScikit-learnTensorFlowOpenCVRasterIOGEEDjangoFastAPIDockerPostgreSQLLangChainGenAINLPCUSUM Algorithm
·

Hire this person

Send a job offer directly to this candidate

PythonSQLNumPySeabornMatplotlib

About

Led architecture and development of a full-stack GI artisan e-commerce platform using Next.js, React, TypeScript, and Supabase. Built a FastAPI, LangChain, and ChromaDB RAG chatbot grounded in Indian GI product and women artisan research, with multi-turn conversational memory. Implemented role-based access control across customer, artisan, admin, and delivery workflows, with Stripe and Razorpay payment integration.

Experience

AI/ML Engineer & Technology Lead at KalaStree (Present)

  • Led architecture and development of a full-stack GI artisan e-commerce platform using Next.js, React, TypeScript, and Supabase.
  • Built a FastAPI, LangChain, and ChromaDB RAG chatbot grounded in Indian GI product and women artisan research, with multi-turn conversational memory.
  • Implemented role-based access control across customer, artisan, admin, and delivery workflows, with Stripe and Razorpay payment integration.
  • Designed evidence-grounded AI retrieval and production API workflows connecting research intelligence with direct-to-artisan commerce.

Associate AI Engineer at Devputers (2025-06 – 2026-04)

  • Designed and deployed scalable AI-enabled web applications using Django and FastAPI, implementing secure authentication (JWT) and modular REST API architecture.
  • Developed "Hindi Meaning App (मतलब क्या)", an AI-powered English learning assistant enabling Hindi-to-English translation and contextual explanations using LLMs.
  • Integrated LangChain with OpenRouter APIs to orchestrate conversational AI workflows and manage persistent chat history.
  • Implemented dual interaction modes: real-time translation engine and conversational teacher mode with contextual examples.
  • Integrated voice input (speech-to-text) and Hindi text-to-speech (gTTS) for interactive language learning.
  • Designed structured conversation memory handling using LangChain message history for session persistence.
  • Developed real-time communication features using Django Channels and WebSockets, enabling live system updates.
  • Architected PostgreSQL database schemas with optimized indexing and query tuning for high-performance systems.
  • Containerized backend services using Docker and docker-compose, enabling consistent development and deployment.
  • Implemented API features including pagination, filtering, rate limiting, and role-based access control.

ML Researcher at Indian Space Research Organisation (ISRO) (2025-03 – 2025-06)

Internship

  • Implemented deep learning models on satellite-derived ocean datasets, optimizing prediction of ocean patterns using MSE and R² metrics.
  • Preprocessed netCDF satellite data with histogram matching and normalization to standardize distributions across multi-sensor inputs.
  • Extracted spatiotemporal features and engineered input variables (sp_inc_angle, ddm_snr, ws) for model training.
  • Applied Levenberg–Marquardt optimization to fine-tune model parameters for improved convergence and reduced error in oceanic predictions.
  • Validated outputs against historical satellite observations and integrated models into forecasting pipelines, enhancing reliability for climate prediction.

MLOps Quality Analyst at Wipro (2023-03 – 2025-03)

  • Labeled and quality-checked 10,000+ 2D/3D images, achieving 98% dataset accuracy for CV models.
  • Enhanced NLP/GenAI pipelines using NLTK error detection and LangChain prompt optimization, boosting F1 by 11%.
  • Designed evaluation workflows for LLM outputs, reducing hallucinations and improving reliability by 15%.
  • Collaborated with domain experts to reduce annotation rework by 25% and improve dataset delivery compliance by 20%.

Data Scientist at Turingminds (2022-06 – 2023-03)

  • Built ML classification models using Scikit-learn with 87% accuracy.
  • Optimized feature selection and preprocessing pipelines, reducing training time by 30%.
  • Developed deep learning models for NLP and time series forecasting, improving accuracy by 18%.

Education

M.Tech in Artificial Intelligence & Data Science – Indian Institute of Technology, Patna (2024-07 – 2026-07)

B.Tech in Mechanical Engineering – Sagi Ramakrishnam Raju Engineering College, Bhimavaram (2018 – 2022)

Skills

AI/ML Engineer & Technology Lead

Reviews

Related