Sai Thummalapudi
iOS Developer
6+ years ML exp
Swift
Python
C
Objective-C
Data Pipeline & Feature Engineering: 4/10
Active 8 days ago
Invite to interview
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
iOS engineer (Middle) specializing in building small to medium Swift and SwiftUI apps with networked features and device integration. The strongest proven skill is writing testable async networking and view models as shown in WeatherApp/WeatherApp/NetworkManager/WeatherAPIManager.swift and WeatherApp/WeatherApp/ViewModel/WeatherViewModel.swift together with WeatherViewModel tests. The work lacks evidence of production release engineering, offline-first sync engines, advanced lifecycle recovery, and measured performance optimizations.
Technical skills
Languages
4
Swift
Python
C
Objective-C
AI/ML
16
PyTorch
TensorFlow
NLP
NumPy
Pandas
Scikit-learn
Statsmodels
LLM
LoRA
Transformers
Computer Vision
LiteRT
PEFT
Sentiment Analysis
Hugging Face
Jupyter Notebook
Mobile
10
Core ML
Firebase
SwiftUI
UIKit
Core Location
XCTest
TestFlight
AVFoundation
Combine
Core Data
DevOps
4
CI/CD
GitHub
GitHub Actions
Git
Other
16
Matplotlib
Seaborn
Swift Concurrency
Firestore
Postman
Classic ML
Dependency Injection
Embeddings
Fine-tuning
MVVM
Edge AI
Recommender Systems
Clean Architecture
State Management
Time Series Forecasting
MVC
Timeline
iOS Developer
•
Middle
Walmart
•
Full-Time
Developed iOS applications using SwiftUI and UIKit with a focus on performance and accessibility. Integrated AI features for image recognition and personalized product recommendations using Core ML and Vision with on-device inference to reduce latency. Built REST API integrations with enterprise backend services and collaborated on converting TensorFlow/PyTorch models into Core ML for iOS deployment. Participated in Agile sprints, code reviews, and TestFlight releases, improving engagement metrics by 25%.
Core ML
Computer Vision
TensorFlow
PyTorch
TestFlight
Agile
University of Maryland, Baltimore County (UMBC)
Master's Degree •
Data Science
iOS Developer Intern
•
Junior
NVIDIA
•
Internship
Built iOS prototypes demonstrating AI/ML use cases including image classification and NLP-driven insights. Optimized deep learning models for mobile and edge deployment, targeting low memory usage and faster inference. Assisted in converting PyTorch models to Core ML and TensorFlow Lite for iOS integration, and used Xcode Instruments to identify and fix CPU and memory bottlenecks. Supported CI/CD pipelines and participated in sprint demos and technical reviews.
PyTorch
Core ML
TensorFlow
CI/CD
Agile
iOS Developer / Software Engineer
•
Middle
Nexova
•
Full-Time
Built internal iOS proof-of-concept apps to visualize analytics and customer insights for stakeholders. Integrated backend APIs to feed data into mobile dashboards and implemented NLP-driven sentiment analysis features connected to iOS UI components. Automated data workflows and supported mobile teams with backend and AI logic while working in Agile teams aligned to business requirements.
Sentiment Analysis
NLP
Agile
Middle AI/ML Engineer
Confidence: Medium Data-centric
Junior data scientist focusing on applied classical machine learning and time-series forecasting with notebook-driven analyses. The strongest proven skill is time-series analysis and forecasting using statsmodels ARIMA together with stationarity testing (Naturalgas_prediction.ipynb shows adfuller_test and ARIMA model fitting). The work lacks productionization, experiment tracking, robust validation pipelines, and any custom model architecture or deployment artifacts.
Model Architecture & Training
3/10
How well models are designed and trained
Uses standard ML algorithms from scikit-learn and classical forecasting via statsmodels ARIMA with straightforward fit/predict workflows; no custom architectures, training loops, or advanced tuning strategies are present.
Evidence
Diabetes Prediction(Logistic Regression).ipynb: log = LogisticRegression(); log.fit(X_train,y_train); pred = log.predict(X_test)
Naturalgas_prediction.ipynb: model = ARIMA(df.Price, order=(1,1,2)); model_fit = model.fit()
Gold_prediction.ipynb: dsr.fit(X, Y); rfr.fit(X, Y) and plot_learning_curve(clf, title) function
Data Pipeline & Feature Engineering
4/10
How data is prepared for models
Typical exploratory data preparation and basic feature extraction are implemented (date parsing, resampling, missing-value imputation, simple column selection); pipelines are notebook-based and ad-hoc rather than production-grade.
Evidence
Naturalgas_prediction.ipynb: df['year'] = pd.DatetimeIndex(df['Date']).year; df.set_index('Date', inplace=True); df.resample('1M').mean()
Naturalgas_prediction.ipynb: df['Price'].fillna(df['Price'].mean(), inplace=True)
Diabetes Prediction(Logistic Regression).ipynb: df.drop('Pregnancies',axis=1,inplace=True); X=df.drop('Outcome',axis=1); y=df['Outcome']
Experimentation & Evaluation
3/10
How results are measured and tested
Contains basic evaluation and validation steps (train/test split, accuracy, classification report, r2_score, cross_val_score and learning curves) but lacks experiment-tracking, reproducible pipelines, or systematic ablation studies.
Evidence
Diabetes Prediction(Logistic Regression).ipynb: from sklearn.metrics import accuracy_score; accuracy_score(y_test,pred); classification_report(y_test,pred)
Gold_prediction.ipynb: scores = cross_val_score(rfr, X, Y, cv = 10); plot_learning_curve(dsr, "Decision_Tree_Regressor")
Naturalgas_prediction.ipynb: model_fit.summary(); residuals.plot(title="Residuals")
MLOps & Deployment
1/10
How models are shipped to production
Minimal to no MLOps or deployment artifacts; models are trained and used for single-run predictions inside notebooks but there is no model serialization, serving, versioning, or CI/CD evidence.
Evidence
Diabetes Prediction(Logistic Regression).ipynb: pred1=log.predict([[85,65,27,0,25.3,0.356,30]]) (ad-hoc inference call inside notebook)
Computational Efficiency
1/10
How efficiently computing resources are used
No GPU/memory optimizations, quantization, or distributed training; only minor use of parallelism (n_jobs) in learning_curve which is typical for notebook experiments.
Evidence
Gold_prediction.ipynb: learning_curve(..., n_jobs=-1) in plot_learning_curve function
Research Depth & Innovation
1/10
Depth of research and new ideas
Work is applied and classical rather than research-oriented; uses well-known statistical tests and models (ADF test, ARIMA) but no novel algorithms, papers reproduced, or custom layers are implemented.
Evidence
Naturalgas_prediction.ipynb: def adfuller_test(price): ...; result = adfuller(df.Price.dropna())
Naturalgas_prediction.ipynb: model = ARIMA(df.Price, order=(1,1,2)); model_fit = model.fit()
Industries
Energy & Utilities• Middle
Financial Services• Middle
Health Care• Middle
Technologies
Classic ML
Firestore
LoRA
GitHub Actions
Fine-tuning
Embeddings
Computer Vision• since 2024
NLP• since 2020
PEFT
CI/CD• since 2024
Transformers
TensorFlow• since 2024
Git
PyTorch• since 2024
LLM
Sentiment Analysis• since 2020
Hugging Face
GitHub
Edge AI
Recommender Systems
LiteRT
Machine Learning• mentioned only
Prophet• mentioned only
Recommendations
- Develop prototype time-series forecasting services and APIs that include model serialization and basic inference endpoints (convert notebook workflow into a script/module and add model save/load).
- Add experiment tracking and reproducible pipelines (use MLflow or Weights & Biases, add parameterized training scripts and deterministic seeds).
- Strengthen evaluation rigor by adding proper cross-validation, holdout strategies, metric dashboards, and simple unit tests for data transformations.
- Practice deploying a small model as a REST service or Streamlit app to demonstrate end-to-end MLOps basics and latency/throughput considerations.
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
A middle-level applied ML practitioner focused on hands-on predictive modeling and time-series analysis using Python notebooks. The strongest proven skill is building end-to-end models and diagnostics as shown by the Naturalgas_prediction.ipynb ARIMA work and the Diabetes Prediction(Logistic Regression).ipynb classification pipeline. There is little evidence of productionization, testing, environment pinning, or engineering practices needed for robust deployment.
Statistical Rigor
4/10
Correct use of statistics
Basic statistical checks are present (ADF test, skew/kurtosis, classification report) but hypothesis framing, uncertainty quantification and rigorous significance/assumption discussions are limited.
Evidence
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Naturalgas_prediction.ipynb: adfuller_test function and printed p-value
Diabetes-prediction/Diabetes Prediction(Logistic Regression).ipynb: classification_report usage for model evaluation
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Gold_prediction.ipynb: skew/kurtosis printed for target distribution
Data Wrangling & Cleaning
4/10
Preparing and cleaning data
Evidence of standard cleaning steps - parsing dates, filling NaNs, dropping columns and simple resampling - but limited treatment of outliers, feature transformations, scaling, and explicit leakage prevention.
Evidence
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Naturalgas_prediction.ipynb: df['Price'].fillna(df['Price'].mean(), inplace=True) and df.set_index('Date') / resample('1M').mean()
Diabetes-prediction/Diabetes Prediction(Logistic Regression).ipynb: df.drop("Pregnancies", axis=1, inplace=True)
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Crudeoil_prediction.ipynb: df['percentChange'].fillna and df['change'].fillna
Exploratory Analysis & Visualization
5/10
Exploring and visualizing data
Multiple exploratory plots and correlation analyses are used to inspect relationships, but written interpretation after visuals is limited and some plots are used without deeper hypothesis-driven commentary.
Evidence
Diabetes-prediction/Diabetes Prediction(Logistic Regression).ipynb: sns.countplot and sns.pairplot usage
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Gold_prediction.ipynb: heatmap of correlations and multiple jointplots
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Naturalgas_prediction.ipynb: lineplots, scatter, pairplot and ACF/PACF plots
Predictive Modeling
4/10
Building models that predict
Uses a sensible baseline-first and multiple model types (logistic regression, KNN, decision trees, random forest, ARIMA) with train/test splits and cross-validation, but lacks hyperparameter tuning rigor, calibration, advanced CV schemes, and deeper error analysis.
Evidence
Diabetes-prediction/Diabetes Prediction(Logistic Regression).ipynb: LogisticRegression().fit and accuracy_score(y_test, pred)
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Naturalgas_prediction.ipynb: ARIMA model.fit and residuals analysis
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Gold_prediction.ipynb: cross_val_score and plot_learning_curve implementations
Business Insight & Impact
2/10
Turning analysis into business value
Some domain-level commentary is present (drivers of commodity prices, identification of features), but there is no mapping to business metrics, no error-cost analysis, and limited actionable recommendations tied to stakeholders.
Evidence
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Naturalgas_prediction.ipynb: textual section listing factors affecting natural gas prices
Diabetes-prediction/README.md: dataset and feature list documented but no cost-of-error or business impact analysis
Reproducibility & Notebook Hygiene
3/10
Clean, repeatable analysis
Notebooks are runnable and sometimes use random_state and parse_dates, but there is no pinned environment, no data/version management, no test suite, and notebooks are not refactored into reproducible scripts or pipelines.
Evidence
Diabetes-prediction/Diabetes Prediction(Logistic Regression).ipynb: train_test_split(..., random_state=42)
An-Intelligent-Agent-Used-to-Predict-the-Queer-Changes-in-the-Prices-of-Various-Commodities/Naturalgas_prediction.ipynb: df = pd.read_csv('natural_gas.csv', parse_dates=['Date']) and resample calls
Industries
Energy & Utilities• Middle
Health Care• Middle
Technologies
Python• Middle
Jupyter Notebook
Scikit-learn
Seaborn
Matplotlib
Pandas
NumPy
Statsmodels
Time Series Forecasting
Machine Learning• mentioned only
Prophet• mentioned only
Recommendations
- Add reproducibility artifacts - requirements.txt or environment.yml, a script-based pipeline (notebook -> module), and model serialization (joblib/pickle) for deployment.
- Introduce robust preprocessing - explicit scaling, outlier handling, feature engineering, and leakage checks with documented rationale and unit tests.
- Strengthen model evaluation - proper CV strategies (time-series CV where applicable), hyperparameter tuning, calibration, and confusion-matrix-driven cost analysis for business impact.
- Refactor notebooks into modular code, add automated tests, and include data/versioning (DVC or similar) to make experiments auditable and production-ready.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Mobile Developer
Confidence: Medium iOS Engineer
iOS engineer (Middle) specializing in building small to medium Swift and SwiftUI apps with networked features and device integration. The strongest proven skill is writing testable async networking and view models as shown in WeatherApp/WeatherApp/NetworkManager/WeatherAPIManager.swift and WeatherApp/WeatherApp/ViewModel/WeatherViewModel.swift together with WeatherViewModel tests. The work lacks evidence of production release engineering, offline-first sync engines, advanced lifecycle recovery, and measured performance optimizations.
Platform Native Mastery
3/10
Knowing the mobile platform
Shows familiarity with modern Swift concurrency and lifecycle hooks but lacks advanced lifecycle handling (process-death, state restoration) and cancellation tied to view lifecycle.
Evidence
WeatherApp/WeatherApp/ViewModel/WeatherViewModel.swift: use of @MainActor and async/await in fetchWeather
WeatherApp/WeatherApp/View/WeatherView.swift: Task usage in onAppear and Task for search
iTunesArtists/iTunesArtists/SceneDelegate.swift: scene lifecycle methods present but left as template stubs
Mobile UI/UX & Responsiveness
3/10
Smooth mobile experience
Reasonable SwiftUI and UIKit UI work with componentization and testable views, but no strong evidence of adaptive layouts, accessibility/dynamic type handling, or advanced multi-device adaptations.
Evidence
WeatherApp/WeatherApp/View/WeatherView.swift: modular SearchBar and WeatherCard SwiftUI components
iTunesArtists/iTunesArtists/View/ArtistTableViewCell.swift: main-thread image set and UI styling in setData and borderUI
Swift_app_Images_to_cloud/The_ultimate_photo/ContentView.swift: SwiftUI UI for image selection and camera buttons
Performance & Battery
1/10
Speed and battery use
Minimal performance or battery-focused work; basic correct network handling but no profiling, baseline measurements, or doze/background battery considerations.
Evidence
WeatherApp/WeatherApp/NetworkManager/WeatherAPIManager.swift: HTTP response validation and JSON decoding
iTunesArtists/iTunesArtists/View/ArtistTableViewCell.swift: image loading with DispatchQueue.main for UI update
Offline & Data Sync
1/10
Working offline and syncing
No offline-first architecture or sync engine; network-only data fetching without local caching, migrations, or idempotent retry strategies.
Evidence
iTunesArtists/iTunesArtists/Network/Network Manager.swift: fetchDataFromAPI performs direct network calls without caching or retry
WeatherApp/WeatherApp/NetworkManager/WeatherAPIManager.swift: direct fetchWeather returning decoded network model without local persistence
Device Integration
3/10
Using device features
Clear device integration experience (camera, photo library, Core Location, Firebase Storage) but missing full permission flows and denial handling.
Evidence
Swift_app_Images_to_cloud/The_ultimate_photo/ContentView.swift: camera invocation and uploadImage using Firebase Storage.storage().reference()
Swift_app_Images_to_cloud/The_ultimate_photo/ImagePicker.swift: UIImagePickerController integration and coordinator delegate
WeatherApp/WeatherApp/View/WeatherView.swift: uses LocationServiceProtocol and MockLocationService for location-based weather fetch
Release & App Lifecycle
2/10
Building and publishing apps
Has unit and UI tests and basic lifecycle hooks, but no visible CI/CD, signing/release configs, staged rollout or crash-reporting integration.
Evidence
WeatherApp/WeatherAppTests/WeatherViewModelTests.swift: async unit tests for WeatherViewModel with mocks
iTunesArtists/iTunesArtistsUITests/iTunesArtistsUITests.swift: UI test scaffolding and launch performance test
Verified artifacts
Expertise
iOS• Middle
Mobile QA & Automated Testing• Junior
Mobile UI/UX• Junior
Technologies
Swift• since 2024 • Middle
Objective-C
UIKit
SwiftUI
Combine
Core ML• since 2024
Core Data
MVVM
Firebase
XCTest
Swift Concurrency
Clean Architecture
MVC
Dependency Injection
State Management
Core Location
AVFoundation
TestFlight• since 2024
iOS• mentioned only
Recommendations
- Develop small to medium iOS apps that require networked features and device integrations such as photo uploads and location-aware content using Swift concurrency and DI for testability.
- Implement an offline-first sync layer with local persistence and idempotent retry to gain experience in robust data sync and conflict resolution.
- Add full permission and denial flows for camera and location and integrate user-facing error handling and settings redirects.
- Build CI/CD and release pipelines (fastlane/CodePush/App Store Connect automation) and wire a crash-reporting tool to demonstrate production readiness.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: