AI & Machine Learning Engineer (GenAI)

Amutha Valli U

638-***-**** ********@*****.*** linkedin.com/in/amutha-valli-u github.com/uammu

EDUCATION

Amrita School of Business, Amrita University July 2022 – April 2024 Master of Business Administration, Business Analytics Coimbatore, Tamil Nadu St. Xavier's College June 2018 – May 2021

Bachelor of Commerce, Computer Application Tirunelveli, Tamil Nadu EXPERIENCE

Generative AI Intern 2026 – Present

Infotact (Project: OmniSight — Multimodal UI Self-Healing & RPA Agent) Remote

● Building OmniSight, a multimodal UI self-healing and RPA agent combining Playwright browser automation with a vision- language model (VLM) for visual QA.

● Developing a FastAPI-based CI/CD gateway to integrate automated visual QA checks into deployment pipelines.

● Applying Generative AI methods as part of the ongoing internship, working iteratively (week-by-week) on the automation and visual-QA components of the agent.

Machine Learning Engineer 2026 – Present

Nexlyra Engineering Works Remote

● Building ML pipelines for real estate and lending risk use cases, including tenant default risk & rent yield optimization, AVM data poisoning & toxic commercial loan default prediction, and structural fatigue sensor-fraud detection (see Projects).

● Working across regression, classification, feature engineering, and model explainability (SHAP) on large-scale (1M-row) datasets.

● Currently ongoing, with models progressing through tuning, evaluation, and deployment-readiness stages. Data Science & Machine Learning Intern 2026

Infotact Remote

● Completed a Data Science & Machine Learning internship at Infotact, building foundational ML and data pipeline skills ahead of transitioning into the Generative AI Intern role at the same company. HR Intern July 2025 – Oct 2025

Sureti Imf Pvt Ltd Coimbatore, Tamil Nadu

● Collected and organized structured candidate information through phone screenings and preliminary evaluations, ensuring completeness and accuracy of applicant details.

● Maintained and updated recruitment databases and spreadsheets with candidate information, performing regular checks to ensure consistency across records.

● Tracked interview progress, outcomes, and follow-up status using structured records, supporting smooth coordination and timely hiring decisions.

Junior Software Developer Intern June 2024 – Dec 2024 Pikture Associates Chennai, Tamil Nadu

● Developed and executed SQL queries to extract, filter, and manage structured datasets from relational databases, supporting data retrieval and analysis requirements.

● Worked extensively with SQL Server to handle structured data, gaining hands-on experience in data storage, querying, and data integrity management.

● Built basic web pages to display application data, improving understanding of data flow between databases and user interfaces. Finance & Data Analytics Intern May 2023 – June 2023 SPR & CO Chennai, Tamil Nadu

● Performed data validation and reconciliation on government housing tender datasets for Tamil Nadu Police Housing Corporation

(TNPHC), reviewing payment vouchers, quotations, and completion reports to ensure completeness, consistency, and compliance.

● Executed KYC data verification and cleansing for ITC retail outlets, validating dealer information, shop name alignment, GST and PAN records, and bank account details through telephonic confirmation to improve vendor data reliability.

● Developed Excel-based analytical and training presentations for corporate users from Larsen & Toubro (L&T), focused on data organization, reconciliation, and reporting fundamentals. PROJECTS

Phantom Structural Fatigue & ML Telemetry Defense Aug 2026 – Present Python, PyTorch/TensorFlow, XGBoost, Optuna, FastAPI, SHAP, Pandas/NumPy, Scikit-Learn/SMOTE, LightGBM

● Serving as Elite Lead Machine Learning Engineer on a structural health monitoring (SHM) system modeling Remaining Useful Life

(RUL) for 50-story buildings from vibration, acoustic, and sway sensor telemetry, designed to flag sensor-spoofing fraud where a single sensor is faked to trigger unneeded emergency retrofit billing.

● Engineered a 51-task Day 1 pipeline to parse nested, chunked SHM JSON, hex-decode sensor array IDs, and interpolate missing RUL targets without lookahead leakage, processing a 1,000,000-row dataset across 40 chunks (~8.24 MB/chunk) with float32 downcasting for memory efficiency.

● Cleaned and validated telemetry labeled ~95% genuine vs ~5% spoofed sensor readings, eliminating Target_RUL_Days nulls and unparseable timestamps down to zero ahead of model development. AVM Data Poisoning & Toxic Commercial Loan Default Prediction Aug 2026 Python, Pandas/NumPy, Scikit-Learn/SMOTE, XGBoost/LightGBM, SHAP, FastAPI

● Parsed and repaired corrupted appraisal data, decoding Base85- and hex-encoded fields and malformed JSON columns as part of a 36-task data preparation pipeline.

● Applied NLP techniques (TF-IDF) on unstructured appraisal notes and combined them with numeric features via hstack to build a unified feature set.

● Built a preprocessing pipeline with memory profiling/downcasting, IterativeImputer, target/one-hot encoding, and PowerTransformer via a ColumnTransformer, saving reusable preprocessor and TF-IDF artifacts.

● Developed a deep learning regressor to predict property valuation (True_Valuation_USD) and an XGBoost classifier to detect toxic commercial loan defaults (Is_Toxic_Default), with SHAP explainability. Tenant Default Risk & Rent Yield Optimization Aug 2026 Python, XGBoost, LightGBM, SQL, Power BI

● Built a regression and classification ML pipeline on a 1M-row dataset to predict optimal monthly rent (Optimal_Monthly_Rent) and tenant eviction risk (Is_Eviction_Risk).

● Performed feature engineering, imputation, encoding, and scaling via a ColumnTransformer pipeline, and conducted feature selection using Pearson/Point-Biserial correlation, VIF, mutual information, and PCA.

● Trained and compared multiple regression (Linear/Ridge/Lasso/Random Forest/XGBoost/Huber) and classification (Logistic Regression/Random Forest/LightGBM/XGBoost) models, applying SMOTE to address class imbalance.

● Tuned final models via RandomizedSearchCV and Optuna, deployed as a 5-page Streamlit dashboard (KPI overview, rent predictor, eviction risk checker, model insights, data explorer) with SHAP explainability, achieving a tuned Rent RMSE of 810.84 and a Risk model AUC of 0.9888 on 1,000,000 records. Geospatial Real Estate Valuation via Spatial Embeddings/GNN July 2026 Python, PyTorch Geometric, XGBoost, Folium

● Cleaned the King County House Sales dataset (21,613 rows), handling missing values and removing outliers via IQR on price to produce a 20,291-row training set.

● Built an ML pipeline to predict property prices, engineering features such as house age, renovation status, and Haversine distance to city center; trained an XGBoost baseline achieving MAPE 16.56%.

● Constructed a K=10 nearest-neighbor spatial graph (122,348 edges) and derived 6 spatial embedding features per property

(neighbor avg/median/std price, distance) to capture location-driven price dependencies.

● Trained and tuned a GPU-accelerated Graph Neural Network, achieving MAPE of 14.69%, outperforming the XGBoost baseline

(MAPE 16.56%) by 1.87 percentage points.

Predictive Maintenance IoT Edge AI System June 2026 Python, LightGBM, SHAP, SMOTE

● Built an ML pipeline on the AI4I 2020 industrial IoT dataset to predict equipment failure, engineering rolling-window and externally fused sensor features to expand the dataset from 14 to 39 columns.

● Applied stratified 5-fold cross-validation with SMOTE oversampling (sampling_strategy=0.5) and hyperparameter-tuned LightGBM to address class imbalance, achieving a mean Macro F1 score of 0.8486.

● Used SHAP for model explainability and conducted noise sensitivity testing and Precision-Recall threshold tuning (final threshold: 0.451) to optimize deployment-ready predictions.

● Diagnosed and resolved a data fusion issue by replacing random sampling with rank-based feature alignment, converting a negative ablation result into a genuine performance improvement. AI Job Trend Analysis Jan 2025

Excel, SQL, Power BI

● Collected and consolidated AI and analytics job market data from multiple sources; performed data cleaning and preprocessing using Excel, including Pivot Tables and Conditional Formatting to highlight trends and demand patterns.

● Analyzed hiring trends, job role demand, and required technical and business skills using SQL queries on structured datasets to identify emerging job opportunities.

● Developed interactive Power BI dashboards to visualize insights and clearly present findings to non-technical stakeholders. Awareness in Cybersecurity – Exploratory Data Analysis April 2024 Excel, Data Visualization

● Designed and conducted exploratory analysis on survey data collected from 50 participants to evaluate cybersecurity awareness and financial risk behavior.

● Cleaned and organized raw survey responses to ensure accuracy and consistency prior to analysis.

● Analyzed response patterns and correlations, summarizing insights using charts and visual representations to support data- driven conclusions.

TECHNICAL SKILLS

Languages: Python, MySQL, JavaScript, HTML/CSS, Power BI, Excel, R Frameworks: ReactJS, PyTorch Geometric, PyTorch, TensorFlow Developer Tools: Visual Studio, DBeaver, Git, Google Colab Libraries: pandas, NumPy, Matplotlib, Seaborn, LightGBM, XGBoost, SHAP, imbalanced-learn (SMOTE), Folium, FastAPI, Playwright, Optuna, Streamlit

Contact this candidate