Headline

Data Scientist & Gen AI Engineer | ML Intern @ CBR, IISc

About

I build AI systems that move from prototype to production across healthcare research, enterprise software, and intelligent backends. Recently completed my MSc in Data Science at VIT Vellore, I've worked across four organizations in under 18 months: modeling neurodegenerative risk at IISc's Centre for Brain Research, engineering AI backends at IIT Ropar, building production AI pipelines at Caterpillar Inc., and applying data science at ServCrust. My core toolkit spans LLM APIs, Amazon Bedrock, Gen AI, Python, and statistical modelling. My BSc in Applied Statistics & Analytics from NMIMS gave me a rigorous mathematical foundation but what drives me is applying that foundation to real systems that create measurable impact. Whether it's reducing risk in clinical data or accelerating decision pipelines in enterprise tools, I care about outcomes, not just outputs. What I bring to a team: a bias for experimentation, comfort working across the full ML lifecycle, and the ability to communicate results clearly to both technical and non-technical stakeholders. 📌 Open to Data Scientist, ML Engineer, and AI Engineer roles (full-time, hybrid/remote). Let's connect: srivally.g29@gmail.com

Location

Greater Bengaluru Area

AI stack

claude code

deepseek

gemini

github copilot

hugging face

langchain

Work experience

Machine Learning Intern at Centre for Brain Research (CBR) IISc, May 2026 - Present · 4 mos, Bengaluru, Karnataka, India, Selected for 4th Edition of CBRAIN Program among 4k+ applicants. Engineered high-dimensional proteomics datasets (300+ features) handling sparse matrices, batch effects, and missing data patterns across multiple time points- production-ready for downstream modelling. Applied ANOVA, survival analysis, and mixed-effects model coefficients to study protein-level changes over time across stable vs. progressive patient groups, identifying statistically significant biomarkers for dementia risk classification. Reduced feature space for dementia risk classification using variance filtering and mixed-effects model coefficients, retaining maximum predictive signal with minimal dimensionality. Building Linear Mixed-Effects Models on longitudinal patient data to model disease progression trajectories and enable early-stage risk stratification. Developing end-to-end ML pipelines in Python (scikit-learn, TensorFlow) with structured validation, reproducible workflows, and clean documentation. Working directly with domain experts to translate complex statistical outputs into clear, decision-ready insights for non-technical stakeholders.

AI Systems/Backend at Indian Institute of Technology, Ropar, Jan 2026 - Apr 2026 · 4 mos, Remote, Selected via NPTEL Topper recognition Reverse-engineered hidden Agmarknet.gov.in government APIs via browser network interception, crafting custom HTTP headers to bypass 403 bot-protection layers and enable live agricultural mandi price data ingestion. Built a dynamic translation engine with fuzzy matching, acronym expansion, de-pluralization, and a zero-dependency date sanitizer supporting 8+ input formats — eliminating all string-to-ID mapping failures. Designed a modular Python backend (core fetcher + state wrappers) and built two FastMCP endpoints for granular and historical price queries, integrated into the Ajrasakha agricultural AI chatbot.

AI Engineer at Caterpillar Inc., Jan 2026 - Feb 2026 · 2 mos, Chennai, Tamil Nadu, India, Built an LLM-powered multilingual document translation pipeline using AWS Bedrock foundation models, supporting PPTX, DOCX, PDF, XML, HTML, and JSON formats with full preservation of layout, fonts, images, and hyperlinks. Replaced Azure Translator API dependency with a custom LLM-based architecture using Python (lxml), improving translation transparency, output quality, and pipeline flexibility. Developed a semantic similarity evaluation framework using TKPERT embeddings to validate contextual preservation and formatting fidelity between source and translated documents. Contributed to CAT AI Translate, Caterpillar's enterprise-scale multilingual document translation platform — from pipeline design through automated validation workflows.

Data Science Intern at ServCrust, May 2025 - Aug 2025 · 4 mos, Hyderabad, Telangana, India, Developed Hex2Vec embeddings for 70+ districts (Karnataka, Maharashtra, Andhra Pradesh), enabling spatial feature representation for AI agents and downstream ML tasks. Optimized sales beat planning by integrating the Travelling Salesman Problem (TSP) with profit–distance trade-offs, improving route efficiency by ~30% compared to baseline methods. Designed and codified an AI Agent using data from 25+ relational tables, incorporating commission, logistics, and market dynamics for accurate pricing decisions. Processed and cleaned 10,000+ reverse-geocoded addresses across multiple states, ensuring reliability of spatial datasets for agent workflows. Co-authored a research paper on Weighted TSP with evolutionary algorithms, including embeddings and multi-objective optimization; accepted by Algorithms for Intelligent Systems, Springer

Education

Master of Science - MS, Data Science, Vellore Institute of Technology

Bachelor of Science - BSc, Applied Statistics and Analytics, SVKM's Narsee Monjee Institute of Management Studies (NMIMS)

Public links

github: https://github.com/Srivally-Garimella