Udit Sharma
AI Engineer
Management: Under a year (1-5 people)
Python
SQL
Data Pipeline & Feature Engineering: 5/10
Model Architecture & Training: 4/10
Active 1 day ago
Download CVCV
Overview
Technical skills
Projects
Timeline
Roles
Overview
A pragmatic ML engineer (mid level) focused on applied time-series forecasting and lightweight conversational tooling with a strong strength in feature engineering for domain problems. The clearest proven skill is time-series feature engineering and forecasting pipeline work, evidenced by the ERA5/met-mast merge, sin/cos wind-direction transforms, lag/rolling features and LightGBM training in the wind prediction notebooks. There is limited evidence of production-grade MLOps, experiment tracking, rigorous reproducibility practices, or research-level model innovation in public code.
Technical skills
Languages
2
Python
SQL
AI/ML
12
LangChain
Pandas
NumPy
Scikit-learn
LightGBM
TF-Keras
Streamlit
LLM
CrewAI
TensorFlow
NLTK
Keras
Other
15
AWS
Matplotlib
Django
FastAPI
Seaborn
GitHub
LangSmith
Jupyter Notebook
Rest API
AI Agents
AI/ML
NLP
ETL/ELT
Multi-Agent Systems
Time Series
Projects
Portfolio & News Publishing Platform
Developed a portfolio and news publishing platform to present content and publish updates. Implemented features supporting content creation and publishing for the platform.
Brain Tumor Detection & NLP Suite
Created a combined suite for brain tumor detection and NLP-related processing. The portfolio project targets extracting or analyzing information using NLP components alongside detection capabilities.
Live Market Agent & Multi-Agent Workflows
Developed multi-agent workflows that automate multi-step research and reporting for live market use cases. Focused on orchestrating agent steps to produce structured outputs from research workflows.
Agentic Research Assistant
Built an agentic assistant to support research tasks by running and evaluating agent workflows. Traced and evaluated agent runs to validate outputs and improve reliability using LangSmith.
Timeline
Project Associate – Data Science & AI
·
Middle
Indian Institute of Technology Delhi
·
Full-Time
Built a wind-farm analytics application forecasting wind speed and estimating energy generation with site output insights. Worked with operational and user-uploaded datasets, using Python and SQL to generate cleaned analytics, KPIs, and narrative reports across performance, availability, and loss dimensions. Contributed to automation of analytics workflows and early detection of degradation before failures.
Python
SQL
Bundelkhand University
Bachelor's Degree ·
Mathematics
Generative AI Engineer Intern
·
Junior
NullClass
·
Internship
Developed GenAI prototypes from idea stage to a working demo in a short remote internship. Focused on iterating on prototype functionality and preparing working outputs based on the initial concepts.
Middle AI/ML Engineer
Confidence: Medium ML Engineer
A pragmatic ML engineer (mid level) focused on applied time-series forecasting and lightweight conversational tooling with a strong strength in feature engineering for domain problems. The clearest proven skill is time-series feature engineering and forecasting pipeline work, evidenced by the ERA5/met-mast merge, sin/cos wind-direction transforms, lag/rolling features and LightGBM training in the wind prediction notebooks. There is limited evidence of production-grade MLOps, experiment tracking, rigorous reproducibility practices, or research-level model innovation in public code.
Model Architecture & Training
4/10
How well models are designed and trained
Practical model selection and evaluation with a tuned LightGBM regressor and standard training calls; lacks custom architectures, advanced training loops, or systematic hyperparameter experiment code.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: LGBMRegressor(...) instantiation and model.fit / model.predict and metric calculation (r2, RMSE)
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP (2).ipynb: LGBMRegressor parameters and feature_importances_ extraction
Data Pipeline & Feature Engineering
5/10
How data is prepared for models
Clear, domain-aware feature engineering and time-series preprocessing (merge_asof, circular time features, sin/cos direction transforms, lag and rolling features, dummy sectors); demonstrates practical data pipeline thinking.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: merge_asof of ERA5 and mast, hour_sin/hour_cos, mast_dir_sin/mast_dir_cos, lag and rolling feature creation
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP (2).ipynb: direction_sector function and get_dummies for mast_sector
Experimentation & Evaluation
3/10
How results are measured and tested
Basic evaluation and simple experiment checks (train/test split, R2/RMSE, feature importance, shuffled-label leakage check) but no formal experiment tracking, ablation studies, or reproducible run configs.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: r2_score and rmse computations and feature importance printing
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP (2).ipynb: shuffled R2 leakage test and re-fit without lag features
MLOps & Deployment
2/10
How models are shipped to production
Lightweight deployment/workflow artifacts: Streamlit app and a local GUI that load models, but no model versioning, CI/CD, containerization, monitoring, or robust serving layer.
Evidence
TOOL_BOT_1/app.py: Streamlit-based front end using cached helpers and tool invocation
Service_chatbot/gui.py: load_model('medquad_model.h5') and local GUI for serving predictions
Computational Efficiency
1/10
How efficiently computing resources are used
Minimal evidence of computational efficiency engineering; standard LightGBM usage with library auto-threading but no profiling, batching strategies, quantization, or GPU optimization.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: LGBMRegressor usage and LightGBM runtime info in outputs
Research Depth & Innovation
1/10
Depth of research and new ideas
No novel research contributions or SOTA implementations; work is applied and pragmatic rather than research-grade innovation.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: applied forecasting pipeline and simulation of future predictions
Expertise
Conversational AI & Chatbots· Middle
AI Agents & Agentic Workflows· Middle
Industries
Energy & Utilities· Middle
Technologies
AI/ML
SQL
LangChain
Scikit-learn
AI Agents
NLP
LightGBM
LangSmith
TensorFlow
Pandas
NumPy
Keras· since 2026
AWS
CrewAI
LLM
NLTK
TF-Keras
Streamlit
GitHub
Multi-Agent Systems
Recommendations
- Harden and productionize models: add model serialization/versioning, simple Docker-based serving, and CI/CD or a deployment recipe for the LightGBM model
- Add experiment tracking and reproducible runs (W&B / MLflow or structured experiment scripts) and log hyperparameters and metrics
- Improve robustness: add input validation and exception handling around network calls (yfinance, requests) and file I/O; remove hard-coded local paths
- Increase MLOps maturity for agent tools: add retries/backoff, timeouts, and sanitize/validate external inputs for safety and reliability
- Develop unit/integration tests and small benchmark scripts to reproduce evaluation numbers consistently
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Data Scientist
Confidence: Medium ML Practitioner
ML practitioner at a middle level specializing in time-series forecasting and applied deep learning for prototyping predictive models. The strongest proven skill is time-series feature engineering and iterative LightGBM forecasting as demonstrated by the Wind-Speed-Prediction notebooks which implement merge_asof, lag/rolling features, evaluation and iterative future prediction. There is limited evidence of production-grade engineering such as unit tests, environment pinning, CI/CD, data versioning or robust deployment pipelines.
Statistical Rigor
4/10
Correct use of statistics
Basic statistical reporting with standard metrics and an explicit leakage check is present, but formal uncertainty quantification, hypothesis testing and deeper statistical rigor are absent.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: R2 and RMSE calculation plus shuffled-y_test leakage check
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP (2).ipynb: model evaluation outputs showing R2/RMSE
Data Wrangling & Cleaning
6/10
Preparing and cleaning data
Strong practical data wrangling and feature engineering for time series - merging with nearest timestamps, sin/cos direction transforms, lag and rolling features, and a standalone cleaning script - though some steps are ad-hoc and lack robust validation or provenance metadata.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: merge_asof usage, lag/rolling feature creation, sin/cos direction transforms
Datascientist_approach/Data_Cleaning_Filtering.py: duplicate removal, invalid-value filters, text standardization and enrichment steps
Exploratory Analysis & Visualization
5/10
Exploring and visualizing data
Notebooks include explanatory narrative and diagnostic plots that communicate modeling choices, but visual analysis is largely descriptive and lacks systematic, answer-driven visual storytelling or annotated insights for stakeholders.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: markdown explanation of feature choices and plotting of actual vs predicted
Brain_tumor_classification/Brain_tumor.ipynb: training accuracy and loss plots with brief interpretation
Predictive Modeling
5/10
Building models that predict
Multiple predictive approaches are implemented (LightGBM for forecasting, CNN for image classification) with feature importance and an iterative forecasting loop, but cross-validation, model selection details and deeper error analysis are limited.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: LGBM training, feature importance and iterative future-prediction loop
Brain_tumor_classification/Brain_tumor.ipynb: CNN model architecture, training loop and model.save
Business Insight & Impact
2/10
Turning analysis into business value
Very limited evidence of connecting results to concrete business metrics or costs of errors; documentation focuses on technical improvements rather than impact or decision rules.
Evidence
Wind-Speed-Prediction-using-ERA5-Met-Mast-Data/MCP.ipynb: narrative describing model accuracy but no business-metric mapping or FP/FN cost analysis
Reproducibility & Notebook Hygiene
3/10
Clean, repeatable analysis
Some reproducibility gestures exist such as setting torch seed and saving models, but there is no pinned environment, requirements file, data versioning, CI, or structured pipeline code to ensure end-to-end reproducibility.
Evidence
Datascientist_approach/Transformer_Architecture.ipynb: torch.manual_seed(42) and detailed shape checks
Brain_tumor_classification/Brain_tumor.ipynb: model.save('BrainTumor.h5')
Industries
Energy & Utilities· Middle
Health Care· Middle
Technologies
Jupyter Notebook
Seaborn
Matplotlib
Time Series
ETL/ELT
Recommendations
- Lead prototypes of time-series forecasting models and feature-engineering pipelines for short-to-medium horizon predictions.
- Build and harden ETL and data-quality pipelines (add provenance, validation tests and schema checks) using the existing Data_Cleaning_Filtering patterns as a starting point.
- Develop lightweight model serving prototypes with FastAPI to expose trained models for internal validation and feedback loops.
- Focus on reproducibility improvements: add pinned requirements, docker/conda environment, data-versioning (DVC) and testable pipeline scripts.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Intern Backend Developer
Confidence: Low Generalist
Junior backend developer focused on simple Django web apps and data models. The strongest proven skill is basic Django data modeling and schema evolution as shown in resume/migrations/0001_initial.py and resume/migrations/0002_rename_highschool_date_resume_timelapse_and_more.py. There is little to no public evidence of API design, testing, security hardening, performance tuning, or production-grade reliability patterns.
API Design
How well APIs are designed
Not evidenced in public code
Data Layer & Database
2/10
Working with databases
Basic schema design and evolution via Django migrations is present, including initial model creation and field renames and removals, but no advanced DB work such as indexes, raw SQL, transactions, or tuned queries.
Scalability & Performance
Handling load and speed
Not evidenced in public code
System Architecture
Overall system structure
Not evidenced in public code
Security & Auth
Protecting data and access
Not evidenced in public code
Reliability & Observability
Stability and monitoring
Not evidenced in public code
Expertise
Python· Intern
Technologies
Python· since 2024 · Middle
Rest API
FastAPI
Django
Recommendations
- Develop simple Django CRUD applications and own the model and migration lifecycle for small features.
- Add automated tests and CI for models and views to build confidence and demonstrate test coverage.
- Work on small API endpoints and record explicit API contracts and error handling to grow toward API engineering tasks.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: