Siddhant Gite - Machine Learning Engineer

Overview

Computer-vision-focused developer at a middle level specializing in practical, heuristic fingerprint quality assessment pipelines. The strongest proven skill is applied image-processing for biometric quality checks implemented in quality_assessment.py (Laplacian variance blur detection, brightness/glare/ROI heuristics and a Gabor-filter ridge clarity check). There is little to no evidence of model training, experiment tracking, production MLOps or comprehensive test coverage in public code.

Phone

Timeline

AI & Data Science Intern · Junior

Bosch · Internship

Jan 2026 to Sep 2026 8 Months Nashik In office

Built an end-to-end machine learning pipeline for injector failure prediction using large-scale industrial datasets, including data preprocessing, EDA, feature engineering, and model evaluation. Addressed class imbalance with resampling and threshold optimization, and trained baseline and gradient-boosting models to improve classification metrics. Developed a Flask-based analysis application with data processing and visualizations, and created an interactive production analysis tool using FastAPI, Streamlit, and Django with automated CSV/Excel reporting.

Pandas

NumPy

Scikit-learn

XGBoost

Joblib

Matplotlib

Flask

FastAPI

Streamlit

Django

SQLite

Projects

LLM-Powered RAG Document Assistant

Apr 2026 to Present

Created a retrieval-augmented generation assistant for context-aware question answering over documents. Implemented document chunking and semantic retrieval using embeddings with FAISS for similarity search. Integrated OpenAI’s GPT through LangChain to generate grounded responses and wrapped the experience in a Streamlit interface.

Customer Churn Prediction Model

Jan 2026 to Present

Developed a churn prediction model using customer attributes and a supervised machine learning approach. Built an end-to-end pipeline with preprocessing and feature scaling, then trained and evaluated a logistic regression model. Delivered an interactive Streamlit dashboard to estimate churn probability and visualize predictions.

Technical skills

Languages

5

Python

C++

SQL

Node JS

C

Python

5

Loguru

Beautiful Soup

Django

FastAPI

Flask

AI/ML

13

OpenCV

NumPy

Pandas

Scikit-learn

Pillow

Streamlit

LLM

PyTorch

LangChain

XGBoost

LangGraph

OpenAI

Joblib

Databases

4

PostgreSQL

MySQL

FAISS

SQLite

Analytics

3

Matplotlib

Power BI

Plotly

DevOps

4

GitHub

Git

Rest API

Docker

Other

8

Machine Learning

AI Agents

CI/CD

RAG

Embeddings

Prompt Engineering

Semantic Search

Master Data Management

Middle AI/ML Engineer Confidence: High Generalist

Computer-vision-focused developer at a middle level specializing in practical, heuristic fingerprint quality assessment pipelines. The strongest proven skill is applied image-processing for biometric quality checks implemented in quality_assessment.py (Laplacian variance blur detection, brightness/glare/ROI heuristics and a Gabor-filter ridge clarity check). There is little to no evidence of model training, experiment tracking, production MLOps or comprehensive test coverage in public code.

Model Architecture & Training

1/10

How well models are designed and trained

Minimal model or training artifacts; image-quality logic is heuristic and there are no training loops, loss/optimizer choices or model architecture files.

Evidence

contactless-fingerprint-quality-assessment/quality_assessment.py: check_ridge_clarity (Gabor filter based ridge scoring)

contactless-fingerprint-quality-assessment/quality_assessment.py: quality_gate (orchestration of checks; no model training)

Data Pipeline & Feature Engineering

2/10

How data is prepared for models

Basic data preprocessing and feature extraction for images (grayscale conversion, thresholding, blur/brightness/glare/ROI heuristics); no large-scale pipeline or augmentation system.

Evidence

contactless-fingerprint-quality-assessment/quality_assessment.py: check_roi_completeness (threshold/Otsu-based ROI fraction)

contactless-fingerprint-quality-assessment/quality_assessment.py: check_brightness (grayscale mean brightness calculation)

Experimentation & Evaluation

1/10

How results are measured and tested

Very light experimentation/evaluation artifacts exist (tiny test scripts for checks), but no reproducible experiment tracking, metrics logging, validation splits or ablation studies.

Evidence

contactless-fingerprint-quality-assessment/test_blur.py: imports and invokes check_blur for simple verification

contactless-fingerprint-quality-assessment/test_quality.py: calls quality_gate for batch testing (basic script-level tests)

MLOps & Deployment

2/10

How models are shipped to production

A simple Streamlit app for serving the quality checks is present, which shows basic deployment/demo capability but no production MLOps (no model/versioning, CI/CD, monitoring or containerization).

Evidence

contactless-fingerprint-quality-assessment/quality_app.py: Streamlit app exposing quality_gate and UI elements

Computational Efficiency

1/10

How efficiently computing resources are used

No explicit computational efficiency work; code uses OpenCV/Numpy for per-image processing but lacks batching, GPU/FP16 optimization, or profiling evidence.

Evidence

contactless-fingerprint-quality-assessment/quality_assessment.py: use of cv2.filter2D and numpy operations without performance optimizations

Research Depth & Innovation

1/10

Depth of research and new ideas

Modest applied research signal (use of a Gabor filter for ridge clarity) but no paper implementations, benchmarks, or novel algorithmic contributions.

Evidence

contactless-fingerprint-quality-assessment/quality_assessment.py: check_ridge_clarity implements a Gabor kernel and computes a ridge_score

Expertise

Biometrics & Intelligent Surveillance· Middle

Industries

Cybersecurity· Middle

Technologies

SQL

C++

MySQL

LangChain· since 2026

FAISS

XGBoost· since 2026

Embeddings

Prompt Engineering· since 2026

Git

SQLite· since 2026

PyTorch

LLM· since 2026

RAG· since 2026

Joblib· since 2026

OpenAI

GitHub

Semantic Search

Machine Learning

Recommendations

Develop and harden unit tests and small reproducible evaluation scripts for each quality check (e.g., pytest tests for check_blur, check_ridge_clarity).
Convert heuristic checks into a small benchmarked pipeline and add benchmark scripts to measure robustness across datasets (create fixed test splits and simple metrics).
Add basic MLOps/serving improvements for production readiness: containerization, simple CI, input validation and logging/monitoring for the Streamlit app.
If moving toward ML, add a small supervised/contrastive training experiment and tracking (W&B or MLflow) to learn ridge clarity or segmentation models from labeled data.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

contactless-fingerprint-quality-assessment qode-market-intelligence Churn_prediction

Middle Data Scientist Confidence: Medium Data Engineer

A data engineer-level practitioner focused on small to medium sized data and ML pipelines with a strength in practical data preprocessing and modular tooling. The strongest proven skill is building end-to-end, production-adjacent utilities for ingestion and preprocessing as seen in the Selenium-based BrowserManager and the DataCleaner/DataStorage modules. There is limited evidence of rigorous statistical analysis, model training pipelines, automated reproducibility (CI or DVC) or large-scale distributed system design.

Statistical Rigor

2/10

Correct use of statistics

Statistical rigor is minimal; there are feature extraction and experiment-style notebooks but no formal hypothesis tests, uncertainty quantification, or treatment of multiple comparisons.

Evidence

qode-market-intelligence/analysis/tfidf_vectorizer.py

deeplearninglab/regularization.ipynb

Data Wrangling & Cleaning

5/10

Preparing and cleaning data

Concrete data cleaning and storage patterns are present with a DataCleaner class, parquet save/load utilities and batch testing scripts, showing practical ETL-focused skills.

Evidence

qode-market-intelligence/processing/cleaner.py

qode-market-intelligence/processing/storage.py

contactless-fingerprint-quality-assessment/test_quality.py

Exploratory Analysis & Visualization

4/10

Exploring and visualizing data

Exploratory analysis and visualization are functional and purpose-driven but lightweight; Streamlit UI and plotting utilities deliver interpretable charts and metrics without deep narrative analysis.

Evidence

contactless-fingerprint-quality-assessment/quality_app.py

qode-market-intelligence/visualization/plots.py

Predictive Modeling

3/10

Building models that predict

Predictive modeling is limited to TF-IDF feature extraction and simple rule-based signal generation; no model training pipelines, robust cross-validation or calibration workflows are shown.

Evidence

qode-market-intelligence/analysis/tfidf_vectorizer.py

qode-market-intelligence/analysis/signal_generator.py

Business Insight & Impact

3/10

Turning analysis into business value

Some problem framing and actionable output exist - trading signals and fingerprint capture guidance - but business-metric linkage, cost-of-error analysis and stakeholder impact details are sparse.

Evidence

contactless-fingerprint-quality-assessment/quality_app.py

qode-market-intelligence/analysis/signal_generator.py

Reproducibility & Notebook Hygiene

3/10

Clean, repeatable analysis

Basic reproducibility artifacts are present such as requirements files and modular code; however there is little evidence of pinned environment automation, data versioning, CI, or disciplined notebook-to-module refactoring.

Evidence

contactless-fingerprint-quality-assessment/requirements.txt

qode-market-intelligence/config/settings.py

deeplearninglab/activation.ipynb

Expertise

Data Science· Middle

Industries

Cybersecurity· Middle

Financial Services· Middle

Technologies

OpenCV

Scikit-learn· since 2026

Matplotlib· since 2026

Beautiful Soup

Pandas· since 2026

NumPy· since 2026

Streamlit· since 2026

Pillow

Loguru

Recommendations

Develop computer vision preprocessing and quality modules for biometric systems using the existing OpenCV-based quality_assessment.py as a foundation.
Build end-to-end ingestion and preprocessing pipelines for market intelligence including robust scraping, deduplication, TF-IDF feature extraction and parquet storage.
Package Streamlit dashboards and small operator UIs to expose quality metrics and guidance, and add unit tests around core checks.
Invest in reproducibility by adding pinned environments, CI tests, and basic data versioning for datasets and outputs.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

contactless-fingerprint-quality-assessment qode-market-intelligence deeplearninglab

Middle DevOps Engineer Confidence: Medium Generalist

A middle-level Python developer focused on practical computer vision for biometric image-quality assessment whose main strength is implementing modular heuristics for image checks. The strongest proven skill is building image-quality checks and a composite scoring pipeline as shown by quality_assessment.py with blur, brightness, glare, ROI and ridge clarity functions. There is no public evidence of CI/CD, containerization, cloud deployment, observability, or production hardening in the human-authored code.

CI/CD Pipelines

Automated build and deploy

Not evidenced in public code

Infrastructure as Code

Managing servers with code

Not evidenced in public code

Containerization & Orchestration

Working with containers

Not evidenced in public code

Observability & Monitoring

Watching system health

Not evidenced in public code

Reliability & Incident Response

Keeping systems up

Not evidenced in public code

Cloud & Cost Optimization

Smart use of the cloud

Not evidenced in public code

Industries

Cybersecurity· Middle

Software· Middle

Technologies

Python· since 2025 · Middle

Tkinter· mentioned only

Recommendations

Turn the quality pipeline into a containerized service with a Dockerfile and a simple CI workflow that builds, tests, and publishes an image.
Add unit and regression tests for the image checks using pytest with image fixtures and expected-score baselines to prevent regressions.
Introduce structured logging and basic observability hooks such as metrics exports or OpenTelemetry traces to track quality-score distributions in production.
Calibrate thresholds and weights with an annotated dataset and add evaluation scripts that produce precision/recall and ROC metrics for the pass/fail decision.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

contactless-fingerprint-quality-assessment Milestone-Monitor

Job preferences

Work permits: India
Employment type: Full-Time