Available Now for Full-Time Roles

Snehlata Kumari

AI/ML Engineer · GenAI · MLOps

Built a 93.6%-accurate satellite imagery classifier using Vision Transformers at NIT Bhopal. Specializing in production-grade GenAI pipelines, RAG systems, and scalable MLOps — turning research into deployable systems.

View Projects Download Resume
GitHub LinkedIn LeetCode GFG Email

Technical Skills

Specialized in production-grade GenAI, Computer Vision, and MLOps pipelines

Programming & Core

PythonSQLC MongoDBGit / GitHubREST APIs

Generative AI & LLMs

LLaMA 3.x / MistralGPT-4o / Claude 3.5 RAG PipelinesLangChain / LlamaIndex Vector DBs (FAISS)RAGAS (Evaluation)

Deep Learning & CV

PyTorch / TensorFlowVision Transformers (ViT) CNN ArchitecturesTransfer LearningOpenCV

MLOps & Infrastructure

MLflowDockerAWS (EC2, S3) FastAPI / StreamlitCI/CD Pipelines

Data Science & Analytics

Pandas / NumPyScikit-learn / XGBoost Feature EngineeringStatistical Modeling Matplotlib / Seaborn

Specialized Domains

GeoSpatial AIExplainable AI (XAI) Legal Tech NLPSemantic Search Prompt Engineering

Work Experience

Real-world impact through ML research & production-grade engineering

Software Developer Intern

Owision · Remote, India

Sep 2025 – Dec 2025
  • Engineered scalable backend architectures using FastAPI and Stripe APIs, optimizing financial data pipelines and increasing reporting efficiency by 25% for the Skillioz platform.
  • Designed and validated high-throughput REST APIs serving a growing user base, reducing production-level integration errors by 30%.
  • Orchestrated automated data validation workflows and filtering systems, improving system-wide data consistency by 20%.
  • Containerized backend services using Docker to ensure environment parity across development and staging, cutting deployment cycle time significantly.
  • Collaborated in an Agile Scrum environment across a cross-functional team, contributing to weekly sprints and performing code reviews.
PythonFastAPI Stripe APIDocker PostgreSQLAgile

AI/ML Research Intern

NIT Bhopal · Bhopal, India

Jun 2025 – Jul 2025
  • Developed a Vision Transformer (ViT) architecture in PyTorch for satellite imagery classification — trained on 10,000+ samples and achieved a peak accuracy of 93.6%.
  • Optimized model inference by 20% through Semi-Supervised Learning and statistical refinement, reducing annotation overhead for future datasets.
  • Performed EDA on multi-spectral satellite bands to identify key spectral features, reducing input dimensionality without loss of precision.
  • Authored comprehensive research documentation detailing XAI (Explainable AI) aspects — attention map visualizations used to interpret model decision-making for geospatial classification.
PyTorchViT GeoSpatial AIXAI EDAFeature Engineering

Featured Projects

End-to-end ML systems — from automated legal intelligence to scalable MLOps pipelines

01
Live Legal NLP XAI

LexAnalyze — Legal Intelligence & XAI

Legal jargon hides critical risks from non-experts, while "black-box" models lack the auditability required for compliance.

Built an ML Ensemble (LogReg + RF) delivering 89.6% accuracy with SHAP word-level attribution → plain-English risk flags. Integrated LLaMA3 via Groq for clause simplification through a grounded RAG pipeline deployed on FastAPI.

89.6%Accuracy
SHAPXAI Layer
<1.2sInference
Scikit-LearnGroq LLaMA3 SHAPFastAPIJan–Mar 2025
GitHub Live App
lexanalyze.streamlit.app
Watch Full Demo
02
Live MLOps Drift Detection

Stock MLOps — Drift-Aware Pipeline

Financial models silently fail due to market volatility and temporal data leakage, making static deployments unreliable in production.

Engineered an end-to-end production lifecycle: MLflow experiment registry + Walk-Forward Validation to eliminate leakage → Evidently AI real-time drift dashboard with automated retraining triggers.

MLflowRegistry
EvidentlyDrift UI
DockerCompose
XGBoostMLflow Evidently AIDockerApr–May 2025
GitHub Live App
stock-mlops.streamlit.app
Watch Full Demo
03
Research Preview GenAI RAG

Scientific RAG — GenAI Copilot

Generic LLMs hallucinate when querying technical literature. Researchers need verifiable answers with multi-document source traceability.

Architected a Hybrid Search (FAISS + BM25) pipeline with cross-encoder reranking → Mistral-7B generation with source-linked citations and hallucination risk scoring via RAGAS evaluation.

HybridRetrieval
Mistral7B-Instruct
RAGASEvaluated
LangChainFAISS Mistral-7BRAGASJan–Feb 2026
GitHub
scientific-rag — Streamlit App
Watch Full Demo

Education

Bachelor of Technology — Computer Science & Engineering

Silicon University · Bhubaneswar, Odisha

Sep 2023 – Present  (3rd Year)  |  CGPA: 9.25 / 10.0

Relevant Coursework: Machine Learning, Data Structures & Algorithms, Database Management, Statistics & Probability, Deep Learning, Natural Language Processing
9.25 CGPA / 10

Achievements & Leadership

Competitive recognition, academic distinction, and leadership development

🥈 Runner-Up

ECOHACK Hackathon 2026

Secured 2nd rank among 120+ teams organized by the Ecosocial Club of Silicon University — recognized for an AI-driven sustainability solution demonstrating rapid prototyping skills.

120+ Teams · Silicon University
🎓 Selected

Aspire Leadership Program

Chosen for a fully funded global leadership development program (Aspire Institute, founded at Harvard University), focused on ethical leadership, community impact, and professional growth.

Harvard-Founded · Fully Funded
💡 Scholar

Herbalife Scholar 2025

Awarded a competitive merit-based scholarship recognizing academic excellence, leadership potential, and sustained academic performance (top 5% of cohort).

Merit-Based · Top 5% Cohort
🇮🇳 Cleared

Smart India Hackathon 2025

Cleared Round 1 of Smart India Hackathon 2025, showcasing innovation and collaborative problem-solving at the institutional level among hundreds of entries.

Round 1 · National Level
📋 Event Lead

Developer Summit 2.0 & Circus of Clues

Led the organization of Developer Summit 2.0 and Circus of Clues with 200+ participants, managing logistics, speaker coordination, and team operations end-to-end.

200+ Participants · End-to-End Lead
💻 500+

DSA Problems Solved

Solved 500+ DSA problems on LeetCode and GeeksforGeeks, mastering algorithms and data structures across arrays, trees, graphs, DP, and competitive programming domains.

LeetCode · GeeksforGeeks

Let's Connect

Available from July 2026 for full-time Data Scientist and ML Engineer roles focused on applied AI, research-driven development, and scalable ML systems. Based in Bhubaneswar, India — open to remote opportunities.

Send an Email Download Resume

Typically respond within 24 hours

Email

kumarisnehlata2005@gmail.com

LinkedIn

Snehlata Kumari

GitHub

Snehlata826
LC

LeetCode

23bced40
GFG

GeeksforGeeks

kumarisneh1ryw
Demo