Sarrthak_v1.0

Open for ML engineering roles

Building intelligent systems from data to deployment.

Machine learning engineer focused on applied AI, production data systems, and agentic software that survives real users.

toy_transformer.forward()live attention
Sarrthak0.24builds0.15production-ready0.15AI0.15systems0.15that0.15
Q.K^T
0.490.470.490.440.490.48
softmax
context[0.75, 0.55, 0.62, 0.70]

Sarrthak builds production-ready AI systems that adapt_

adapt26%ship26%scale25%learn24%

Featured Projects

Projects recruiters can scan quickly

The work below emphasizes product scope, architecture choices, and measurable technical value instead of just listing tools.

rnn_cell.forward(h_t, x_t)sequence → word
h0h1h2h3h4h5
decoded →_ _ _ _

GitHub Activity

Contributions (YTD)1.2K
Public Repos42
Private Mode--

Syncing GitHub activity...

Education

M.S. Data Science

2025 - 2027

Indiana University Bloomington

Applied Machine LearningApplied Database TechnologiesComputer Vision

B.Tech Computer Science

2019 - 2023

Vellore Institute of Technology

CGPA 3.8/4

Data MiningNeural NetworksMachine LearningCloud ComputingGitArtificial Intelligence

Recent Achievement

Winner - Luddy Hackathon

Designed a Bloomington Transit app that turns public transit data into live bus tracking, route visibility, arrival estimates, schedules, and commute alerts for students navigating campus.

1st Place$1,000 Prize24 Hr BuildBloomington Transit APIGTFS RealtimeMVVM Architecture
Situation
Luddy Hackathon challenged graduate teams to design a Bloomington Transit System as a real Android transit app using the public Bloomington Transit API, live vehicle data, route maps, schedules, and arrival alerts.
Task
Deliver a working demo in 24 hours with individual bus tracking, 10-second API polling, accurate expected-arrival calculations, an interactive route view, a schedule table, and a clear 3-minute judge presentation.
Action
Worked through the problem statement by connecting Bloomington Transit and GTFS feeds, structuring the app around an MVVM-style architecture, building live map and route surfaces, adding polling-backed vehicle updates, and shaping the demo around reliability and usability.
Result
Won 1st place at Luddy Hackathon and earned $1,000 in prize money.

Experience Timeline

Professional experience

Data Scientist

Indiana University Bloomington

Aug 2025 - Present
  • Situation: Complex Olympic diving records required faster search, review, and classification across large-scale text and video data.
  • Task: Build multimodal retrieval and model-training systems that could support natural language querying, domain classification, and automated scoring.
  • Action: Engineered a Hadoop-backed RAG pipeline with vector embeddings and GLM-4.5 synthesis, fine-tuned LLaMA with LoRA on 10,000 records, and adapted VideoMAE V2 spatio-temporal attention layers for dive biomechanics across historical videos.
  • Result: Improved retrieval accuracy by 40%, reduced manual review by 15 hours per week, increased few-shot prediction accuracy by 60%, and mapped multi-frame videos to an automated grading metric.

Sr. Software Engineer

Optum Global Solutions

Jan 2022 - 2025
  • Situation: Healthcare member platforms needed faster, more reliable product experiences across frontend, backend, claims data, and deployment surfaces.
  • Task: Modernize legacy AEM experiences, improve service response paths, support Kubernetes migration, and turn member experience data into actionable leadership insight.
  • Action: Migrated key widgets to React and Next.js, built Spring Boot services for login/member APIs, helped configure Kafka claims topics, contributed Kubernetes service/proxy files, built KPI dashboards, modeled pain points with Python/Keras/Scikit-Learn, and prototyped LLM features for patient claim forms.
  • Result: Expanded from frontend delivery into full-stack, platform, ML, and LLM production work while supporting high-volume healthcare workflows and sub-50ms response-time goals.

Core Stack

Languages

PythonTypeScriptSQLR

AI / ML

PyTorchTensorFlowLangGraphOpenAIComputer Vision

Frontend

Next.jsReactTailwind CSSAEM

Backend

FastAPISpring BootPostgreSQLRedisNeo4jKafka

DevOps

DockerKubernetesGitHub ActionsCloudflareVercel

Resume

Machine learning engineering resume with project, research, and production software experience.

Contact

Recruiters, collaborators, and hiring teams can reach me directly. The form opens your email client with the message fields prefilled.