AI RESEARCH LAB // ONLINE
Associate Data Scientist @ Trinity Mobility · Bengaluru, Karnataka, India
Data Scientist & AI / ML Engineer with 2.5 years of experience carrying machine learning work from raw problem statements through to production. Specializing in time-series forecasting, multi-stage decision pipelines, synthetic data simulation, and MLOps tooling that bridges data science with robust software delivery.
Gujuri Alekha (also Alekha Gujuri)
ASSOCIATE DATA SCIENTIST
At Trinity Mobility, my work centers on applied machine learning for smart-city systems: designing three-stage automated irrigation pipelines, building smart energy grid simulation engines, integrating ML outputs with Integrated Command and Control Centers (ICCC), and deploying LLM-based alert summarization for operator traceability.
Beyond standalone modeling, I engineer the platform tooling that makes models reliable and maintainable: connecting Model Registries with Nexus, building single-click CI/CD deployment modules, and enforcing automated test quality gates with code coverage in Flask/FastAPI microservices.
Grounding my engineering practice is a strong quantitative foundation: currently completing an MCA specializing in Machine Learning & Artificial Intelligence from Lovely Professional University, preceded by a Bachelor of Science (Physics Honours) from Science Degree College (CGPA 7.7/10).
Trinity Mobility → Agentic AI workflows orchestrating multi-step reasoning across forecasting and decision stages.
Smart Irrigation → Chained 3-stage pipeline: soil & weather forecasting, irrigation ON/OFF classification, duration regression.
Vision projects → Custom CNN for jellyfish species, CIFAR-10 recognition engine, ResNet50 transfer learning.
NLP Suite → Reusable text suite covering tokenization, lemmatization, TF-IDF vectorization and sentiment classification.
Trinity Mobility → RAG pipelines and LLM-based alert summaries so ICCC operators read grounded, plain-language explanations.
Face Mask Detection → Custom-trained YOLOv5 detecting face mask compliance in high-density public environments.
Designing chained predictive pipelines (e.g. Smart Irrigation) where soil and weather forecasts feed decision classifiers and duration regressors, returning unified operational commands.
Building physics-informed synthetic data engines and simulation pipelines (e.g. Smart Energy) for load curves, peak demand, and anomaly scenarios across 15-min to daily intervals.
Integrating RAG pipelines and agentic AI workflows with municipal command centers (ICCC). Deploying LLM-based alert summarization with retrieval-augmented context so operators receive clear, grounded operational summaries.
Architecting Model Registries with Nexus, single-click deployment CI/CD workflows, and automated test execution with code coverage gates within Flask/FastAPI service pipelines.
Grouped by area. Brighter chips are the tools I use most in production.
End-to-end model development spanning supervised, unsupervised, and time-series domains.
Synthetic load generation, grid anomaly scenarios, and LLM-based alert interpretation.
Production inference serving, schema validation, and inter-service telemetry.
Centralized model versioning, automated testing gates, and single-click CI/CD.
High-throughput data transformations, serialization formats, and graph/relational databases.
Four pipelines I build with, each traced to a project that uses it.
Every model I ship follows the same path, from raw sensor or business data to a prediction an operator can act on.
DATA: Raw sensor and weather readings arrive, with gaps and noise.
Illustrative example
Follow one sentence through the stages. Each step lights up in turn.
TOKENIZATION: The sentence is split into tokens, the units a model reads.
Illustrative example
A query is embedded, matched against a vector index, and the retrieved context grounds the LLM so its answer is traceable to real data. It is the pattern behind the alert summaries I deployed for city operators.
USER QUERY: An operator asks a plain-language question.
Illustrative example
Images become pixel tensors, convolutional layers extract features, and a vision model classifies what it sees or localizes it with bounding boxes.
IMAGE: A camera frame arrives.
Illustrative example
Select a project for the full case study.
SMART CITY IOT & TIME-SERIES
Sequential 3-stage predictive pipeline forecasting environmental conditions, classifying irrigation necessity, and regressing duration—exposed via FastAPI with AI-generated alert summaries for city command center operators.
ML ENGINEERING & MLOPS
Enterprise platform enabling non-ML specialists to ingest data from SQL/APIs, clean, train regression/classification/clustering models, enforce automated quality gates, and deploy via a Nexus Model Registry.
SIMULATION & SYNTHETIC DATA
Physics-informed simulation engine and synthetic dataset generator modeling distribution grid loads, peak demands, and electrical anomalies across 15-minute, hourly, and daily resolutions for ICCC dashboards.
COMPUTER VISION & EDGE AI
Automated real-time computer vision system built on custom-trained YOLOv5 to detect face mask compliance in high-density public environments.
Additional machine learning models, computer vision Capstones, and statistical notebooks.
Predictive customer conversion modeling on 45,000+ interactions to maximize term deposit subscriptions while reducing cold-calling operational overhead.
Convolutional neural network for marine organism classification to support ecological surveys and automated taxonomic indexing from underwater imagery.
Deep convolutional architecture extracting robust feature representations from noisy 32x32 color images across 10 balanced classes.
Predictive diagnostic classification model using 10 biochemical blood biomarkers, achieving 82% diagnostic accuracy for early triage.
Most recent first.
Trinity Mobility website (opens in new tab)Bengaluru, Karnataka, India
2024 — Present (2 years 5 months)
Joined in an internship capacity and transitioned into the full-time Associate Data Scientist role. Responsible for carrying machine learning systems from domain problem statements through to production delivery across municipal smart-city platforms.
Rubixe website (opens in new tab)Bengaluru, Karnataka, India
January 2023 — January 2024 (1 year 1 month)
Conducted advanced data analysis, developed predictive machine learning models, and managed end-to-end data engineering pipelines to extract actionable decision intelligence for enterprise clients.
EDUCATION
Lovely Professional University · Feb 2025 — Feb 2027 (In Progress)
Specialization in Machine Learning & Artificial Intelligence
EDUCATION
Science Degree College · Aug 2018 — Sep 2021
Physics Honours & Applied Mathematics · CGPA: 7.7 / 10
EDUCATION
Science College · 2016 — 2018
Physics, Chemistry, Mathematics & Information Technology
CERTIFICATION
NASSCOM & FutureSkills Prime · September 2023
Ministry of Electronics and IT (MeitY), Govt. of India
CERTIFICATION
International Association of Business Analytics (IABAC) · September 2023
Global Professional Standard
CERTIFICATION
DataMites™ · January 2023 — May 2023
DataMites Bangalore Global Training
Hands-on Python notebooks on time series, statistics, and machine learning.
Learn how to identify White Noise and Random Walk processes in financial time series using visual, ACF/PACF, and Ljung-Box test diagnostics.
Learn how to detect non-stationarity in time series using statistical tests and apply techniques like differencing, transformations, and detrending to achieve stationarity.
Learn the fundamentals of time series forecasting, additive vs. multiplicative models, and robust STL decomposition using NIFTY 50 data.
An introduction to Python, its interpreters, and core data types: numbers, strings, lists, tuples, dictionaries and basic operators.
Whether you have an applied machine learning project, an engineering role, or a technical inquiry, my inbox is always open.
