Aditya Boghara
Software Developer
C++
JavaScript
Node JS
C
Swift
Python
SQL
Data Pipeline & Feature Engineering: 4/10
Experimentation & Evaluation: 4/10
Active 14 days ago
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
A practical engineer focused on applied ML and LLM-enabled tooling at a solid middle level. The strongest proven skill is building an end-to-end interactive RiskLens assessment and AI report flow, evidenced by risklens_ai/app.py together with scoring.py, simulator.py and report.py. Public code lacks production MLOps, formal testing, efficiency engineering and novel research-level model work.
Technical skills
C++
JavaScript
Node JS
C
Swift• Middle
Python• Middle
SQL• Junior
C++
PyTorch C++
TensorFlow C++
Python
Flask
Databases
MySQL
PostgreSQL
Oracle
AI/ML
Embeddings
Fine-tuning
Keras
LangChain
LlamaIndex
LLM
LoRA
MLFlow
NLP
NumPy
Pandas
PEFT
Prompt Engineering
PyTorch
QLoRA
RAG
Scikit-learn
Streamlit
TensorFlow
Time Series Forecasting
Tokenization
Transformers
OpenAI SDK
TF-Keras
YOLO
CatBoost
XGBoost
Mobile
iOS
UIKit
ARKit
SceneKit
DevOps
Amazon EC2
CI/CD
Docker
Git
GitHub Actions
Rest API
AWS
Analytics
Plotly
Design
Figma
Timeline
University of Houston
Master's Degree •
Management Information Systems
Instructional Assistant
•
Middle
University of Houston
•
Full-Time
Supported 100+ graduate students with debugging SQL queries, schema design, and database configuration on Windows and macOS. Created standardized Oracle troubleshooting guidance to reduce recurring setup failures and improve onboarding consistency. Reviewed student assignments for data integrity, query correctness, and performance, and provided actionable technical feedback.
SQL
Oracle
Backend Software Engineer
•
Middle
Project Human City
•
Full-Time
Designed MongoDB-backed data models for a web and mobile platform serving 10,000+ users. Implemented a cross-platform authentication system to strengthen security and access control. Produced architecture documentation adopted by multiple teams and improved developer productivity by setting up a GitHub Codespaces and MongoDB Atlas development workflow.
Arizona State University
Bachelor's Degree •
Computer Science
Middle AI/ML Engineer
Confidence: High LLM Engineer
A practical engineer focused on applied ML and LLM-enabled tooling at a solid middle level. The strongest proven skill is building an end-to-end interactive RiskLens assessment and AI report flow, evidenced by risklens_ai/app.py together with scoring.py, simulator.py and report.py. Public code lacks production MLOps, formal testing, efficiency engineering and novel research-level model work.
Model Architecture & Training
3/10
How well models are designed and trained
Standard model-building and tuning present (notebooks, Keras MLP, XGBoost, grid/random search) but no custom architectures or distributed training. Engineering is competent for prototyping and applied ML tasks.
Evidence
MLProject/src/notebook/Model Training.ipynb: build_sequential() Keras MLP and EarlyStopping usage
MLProject/src/components/model_trainer.py: evaluate_models + model selection and hyperparameter grids
Insurance-Classification-Project/Project.ipynb: Keras training and RandomizedSearchCV/XGBoost tuning
Data Pipeline & Feature Engineering
4/10
How data is prepared for models
Concrete data engineering patterns and reusable preprocessing objects are implemented (ColumnTransformer, pipelines, preprocessor saving). Data ingestion and transformation components exist for an end-to-end flow.
Evidence
MLProject/src/components/data_transformation.py: get_data_transformer_object() building ColumnTransformer and pipelines
MLProject/src/components/data_ingestion.py: initiate_data_ingestion() produces train/test CSVs and returns paths
MLProject/src/pipeline/prediction_pipeline.py: PredictPipeline and CustomData for serving predictions
Experimentation & Evaluation
4/10
How results are measured and tested
Reasonable experimentation and evaluation practices appear: cross-validation, CV hyperparameter search, standard metrics and selection logic, but no experiment tracking or reproducible run management (W&B/MLflow) shown.
Evidence
Insurance-Classification-Project/Project.ipynb: RandomizedSearchCV, PR/ROC evaluation and threshold selection
MLProject/src/components/model_trainer.py: evaluate_models() integration and best-model selection
MLProject/src/notebook/Model Training.ipynb: baseline comparisons and metric reporting
MLOps & Deployment
3/10
How models are shipped to production
Basic serving and delivery artifacts exist - Streamlit app for interactive assessment and a Flask app scaffold for prediction - but no production-grade deployment, CI/CD, monitoring or model lifecycle automation shown.
Evidence
risklens_ai/app.py: Streamlit UI with exports, session state and download endpoints
MLProject/app.py: Flask application exposing predict_datapoint endpoint and loading prediction pipeline
Computational Efficiency
1/10
How efficiently computing resources are used
Minimal attention to computational efficiency beyond standard training knobs (batch sizes, n_jobs). No evidence of GPU profiling, quantization, memory optimization or inference acceleration.
Evidence
Insurance-Classification-Project/Project.ipynb: use of batch_size in Keras training and RandomizedSearchCV with n_jobs
MLProject/src/components/model_trainer.py: hyperparameter grids but no GPU/efficiency instrumentation
Research Depth & Innovation
2/10
Depth of research and new ideas
Some thoughtful domain heuristics and a structured LLM prompt architecture are present, but there is no novel algorithmic research, new neural layers or SOTA reproduction. The prompt engineering is the most innovative element.
Evidence
risklens_ai/report.py: detailed system_prompt and user_prompt to control LLM output structure
risklens_ai/scoring.py: deterministic risk and ROI formulae plus dependency/blocking logic used by simulator
Verified artifacts
Expertise
LLM• Middle
Cybersecurity & Threat Intelligence• Middle
Industries
Cybersecurity• Middle
Education• Middle
Financial Services• Middle
Technologies
SQL• Junior
C++
MySQL
PostgreSQL
Rest API
LangChain
CatBoost
Oracle
Flask
LlamaIndex
LoRA
MLFlow
YOLO
GitHub Actions
Fine-tuning
Embeddings
Prompt Engineering
NLP
PEFT
QLoRA
CI/CD
Transformers
TensorFlow
NumPy
Keras
Git
PyTorch
AWS
Docker
LLM
RAG
TensorFlow C++
PyTorch C++
Amazon EC2
Tokenization
Time Series Forecasting
Recommendations
- Develop interactive LLM-driven tooling and assessment UIs (Streamlit) that produce structured reports and exports - expand the existing RiskLens flow into a production-ready web service.
- Build lightweight MLOps around the pipelines: automated model versioning, CI for tests, and simple deployment (Docker + basic monitoring) for the Flask/Streamlit endpoints.
- Deliver ML prototypes and business-facing models in cyber and insurance domains - use the current scoring/simulator design to productize prioritized remediation suggestions.
- Focus on improving experiment tracking and reproducibility (MLflow or W&B) and add unit/integration tests for scoring and simulation logic.
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 pragmatic ML practitioner at a middle level who builds end-to-end analytics products with UI, deterministic scoring engines and simulation capabilities. The strongest proven skill is delivering a domain-focused risk-scoring and simulation pipeline that ties controls to ROI and prioritized remediation (scoring and simulation modules plus Streamlit dashboard). Public artifacts show limited production hardening - there is little evidence of automated tests, CI, pinned environments, or extensive uncertainty quantification.
Statistical Rigor
4/10
Correct use of statistics
Solid deterministic scoring and ROI logic is present but there is minimal statistical inference, no uncertainty quantification, and no formal hypothesis testing or multiple-comparison controls.
Data Wrangling & Cleaning
5/10
Preparing and cleaning data
Practical preprocessing and data modeling are evident (ColumnTransformer, SMOTE, dataclass-based question metadata), but dataset provenance/versioning and robust cleaning pipelines are limited.
Exploratory Analysis & Visualization
6/10
Exploring and visualizing data
Exploratory visuals and narrative are purposeful - Streamlit dashboard includes radar charts, peer comparisons and written change narratives; notebook EDA includes interpretations rather than only decorative plots.
Predictive Modeling
6/10
Building models that predict
Predictive workflow shows baseline-first modeling, SMOTE/class-weight handling, neural nets and XGBoost with hyperparameter search and evaluation (AUC/F1/PR), but test harnessing and robust cross-validation artifacts are moderate rather than industrial-grade.
Evidence
AdityaBoghara/Insurance-Classification-Project/Project.ipynb: TF-Keras model training with early stopping, SMOTE resampling and threshold selection via precision-recall
AdityaBoghara/Insurance-Classification-Project/Project.ipynb: RandomizedSearchCV over XGBoost and evaluation with AUC/F1 and PR analysis
Business Insight & Impact
6/10
Turning analysis into business value
Clear business focus - ROI-ranked actions, org-type and org-size tailored notes, and an action tracker with exports show strong product-to-metric thinking; measurable business impact is explicitly considered.
Reproducibility & Notebook Hygiene
4/10
Clean, repeatable analysis
Some reproducibility hygiene exists (requirements.txt, dotenv usage, seed in notebooks) but there is no evidence of pinned environment lockfiles, CI, test suite, or data versioning for reliable production re-runs.
Expertise
Analytics• Middle
Industries
Cybersecurity• Middle
Financial Services• Middle
Technologies
Python• Middle
XGBoost
Scikit-learn
OpenAI SDK
Plotly
Pandas
TF-Keras
Streamlit
Recommendations
- Lead development of analytics-first product features: risk-scoring engines, ROI simulators and dashboards that tie model outputs to concrete remediation steps.
- Develop ML-to-production hygiene: add automated tests (unit and integration), CI pipelines, pinned environments/lockfiles and simple data versioning so results are reproducible.
- Harden LLM/AI integrations and failure modes: add safe fallbacks, retry/timeouts, clear prompt governance and logging for generated recommendations.
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
An iOS developer at an early-to-mid level with hands-on experience building AR/SceneKit prototypes. The strongest proven skill is AR integration using ARKit and SceneKit, evidenced by image-tracking, plane-detection and hit-test node placement (e.g. ViewController.swift using ARWorldTrackingConfiguration and ARImageTrackingConfiguration). There is little or no evidence of advanced mobile engineering practices such as state restoration, background/offline sync, permission rationale handling, automated testing, or release/CI pipelines.
Platform Native Mastery
2/10
Knowing the mobile platform
Basic platform mastery - correct use of AppDelegate lifecycle and view lifecycle hooks, but no state restoration, process-death handling or structured concurrency.
Mobile UI/UX & Responsiveness
2/10
Smooth mobile experience
Mobile UI/UX shows AR-focused interactive UX and FPS/statistics toggles but lacks adaptive layouts, accessibility, dynamic type, or multi-device design considerations.
Performance & Battery
1/10
Speed and battery use
Minimal evidence of performance or battery engineering - only scene statistics/debug options present, no measurements, profiles, or doze/background-aware work.
Offline & Data Sync
Working offline and syncing
Not evidenced in public code
Device Integration
3/10
Using device features
Strong device-integration with ARKit and SceneKit APIs - plane detection, image tracking, hit-testing and node management are correctly used, but explicit permission flows and denial handling are not implemented.
Release & App Lifecycle
1/10
Building and publishing apps
No release automation, CI, crash reporting, or store/release artifacts are present; only basic AppDelegate lifecycle stubs exist.
Expertise
iOS• Middle
Technologies
iOS
Swift• Middle
UIKit
ARKit
SceneKit
Recommendations
- Lead short-term AR prototype work - build interactive AR demos and product concepts using ARKit + SceneKit.
- Implement small to medium iOS features that require camera/3D integration, hit-testing and node management (e.g., measurement tools, image-triggered experiences).
- Work on iOS UI and interaction polish for AR experiences - add accessibility, adaptive layouts, permission flows and basic performance profiling.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: