Amit Kumar
GitHub LinkedIn
Résumé

Amit Kumar AI/ML & Data Engineering · New Delhi

The unglamorous
half of ML.

I build the parts of an ML system that decide whether anyone can trust it. Most of it is infrastructure, and most of that is unglamorous.

Currently
Machine learning engineer at IronLabs AI, building agent tooling and model routing · Mar 2025 to present
Research
6 peer-reviewed papers, 78 citations · IEEE, ICPR, Springer, Research Square
Focus
LLM systems · retrieval · streaming data platforms · computer vision
Open to
Full-time AI/ML and data engineering roles · also happy to talk research collaboration
See the work Résumé GitHub
Selected work 15 projects · 2023 to 2026

Selected work

Everything else

All projects, most recent first
Year Project What it does Links

Where the work has been

Machine Learning Intern

IronLabs AI · Mar 2025 to present

Agent optimisationMeta-agentModel routingPlugin system
  • Agent-optimisation service that rewrites and improves a user's agent code
  • Meta-agent proposes each change through a propose-benchmark-decide loop
  • Every iteration sandboxed, smoke-tested and traced end to end
  • Model routing behind a unified model gateway, with a Python SDK and a plugin system

Computer Vision Intern

Mahyco · Aug 2024 to Dec 2024

Computer visionDrone imageryDeep learning
  • Deep-learning pipeline for crop-yield prediction from aerial drone imagery
  • Handled occlusions and visual noise in the imagery
  • Cut counting MAE by 25% and improved detection mAP

B.Tech, Data Science & AI

IIIT Naya Raipur · Nov 2022 to Jul 2026

CGPA 8.69DSAI
  • CGPA 8.69 / 10
  • Coursework across machine learning, deep learning, NLP, computer vision, big-data systems and applied statistics
  • The stress-modelling papers below came out of it
Tooling 34 tools · six groups

What I reach for

Languages
Python · C · C++ · SQL
Backend & APIs
FastAPI · Flask · Streamlit
Data
PostgreSQL · Elasticsearch · Kafka · Airflow · Spark · Redis · FAISS
ML & deep learning
PyTorch · TensorFlow · Keras · scikit-learn · XGBoost · OpenCV · Pandas · NumPy · NLTK
LLM engineering
LangChain · LangGraph · MCP · RAG · Pydantic · Pinecone
Ship & observe
Docker · docker-compose · Git · GitHub Actions · MLflow
Publications 6 papers · citations from Google Scholar

Six papers

Peer-reviewed work, mostly stress modelling from wearables and validated surveys, plus a drone-vision framework. Code and datasets are linked wherever they're public.

  1. V-Surveillance: A Hybrid Deep Learning Framework for Real-Time Aerial Surveillance Using Drone Imagery IEEE CICT 2025 · Scholar
  2. Machine Learning Algorithms for Detecting Mental Stress in College Students IEEE I2CT · arXiv · Scholar
    38 citations
  3. Optimizing Well-Being: Unveiling Eustress and Distress Through Machine Learning IEEE ICICT · code · Scholar
    14 citations
  4. AI-Enabled Real-Time Exercise Monitoring with MediaPipe and OpenCV IEEE ICCCNT · DOI
    17 citations
  5. A Deep Learning Approach for Early Stress Detection Using Electrodermal Activity Through Wearable Devices ICPR Workshops, Springer · DOI
    8 citations
  6. A Novel Integrating Multimodel Sensor Data with Machine Learning for Stress Detection Research Square preprint · DOI
    1 citation