Open to Opportunities · Hyderabad, India

Shiva
Kumar
Vanam

AI / Machine Learning Engineer

◈ Anthropic ◈ OpenAI ◈ NVIDIA ◈ Microsoft ◈ Google DeepMind ◈ Meta AI ◈ Apple ML ◈ Cohere ◈ Mistral AI ◈ xAI

Building production-grade LLM systems, RAG pipelines with Constitutional AI safety layers, and GPU-accelerated inference engines. Passionate about alignment research and deploying responsible AI at scale.

✉ Get in Touch ⌥ GitHub ⌥ LinkedIn ⌥ LeetCode ↓ Resume

1500+

Students Mentored

3.2×

Inference Speedup

63%

Harm Reduction

0.93

RAGAS Relevance

Shiva @ Solana APEX Mumbai

Core Strengths

What Sets Me Apart

🧠

Alignment-First Engineering

Built a Constitutional AI critique layer into production RAG — safety metrics alongside performance. Helpfulness and harmlessness must coexist as first principles, not afterthoughts.

⚡

Full-Stack LLM Engineer

From transformer architecture from scratch to GPU inference optimization with TensorRT + Flash Attention — operating across the entire model lifecycle, not just the API layer.

📊

Evaluation Obsessed

Implemented full RAGAS evaluation suites with reproducible benchmarks. You can't improve what you don't measure — and most teams aren't measuring the right things.

🎯

Teaching at Scale

Mentored 1,500+ students across India on AI/ML fundamentals under Prof V S Raju (Retd. IIT Delhi Director). Deep knowledge + ability to communicate complex ideas simply.

Technical Arsenal

Skills & Tools

01 / LLM & RAG

LLM & RAG Systems

Large Language ModelsRAGFAISSLangChainOpenAI APIHuggingFacePrompt EngineeringVector DBs

02 / ALIGNMENT

Agentic AI & Alignment

RLHFPPOConstitutional AILoRA/PEFTTRLAgentic SystemsTool UseFlash AttentionINT8/INT4

03 / EVALUATION

Evaluation & Observability

RAGASLLM Eval FrameworksA/B TestingModel MonitoringAnomaly Detection

04 / MLOPS

MLOps & Deployment

DockerFastAPIFlaskREST APIsAWS EC2FirebaseCI/CDModel Serving

05 / ML ENG

ML Engineering

PyTorchTensorFlowscikit-learnXGBoostNLPComputer VisionDeep LearningTime Series

06 / TOOLING

Programming & Tools

Python (Advanced)SQLC++JavaScriptGitW&BLinuxPostgreSQLTableau

Work History

Experience

Jul 2025 — Present

VSR Consultants Pvt. Ltd

Hyderabad, India

Software Engineer & AI Mentor

Deliver daily AI/ML instruction to 1,500+ students across India at an NGO founded under Prof V S Raju (Retd. IIT Delhi Director), covering supervised learning, neural networks, and LLM fundamentals.
Designed and deployed the official NGO platform and consultancy website from scratch using Firebase Studio, Firestore, and CDN hosting.

Teaching · 1500+ StudentsFirebaseLLM Fundamentals

Jan 2025 — Apr 2025

Infexial Software Solutions

Pune, India

AI / ML Engineer Intern

Productionized a supervised classification pipeline (XGBoost + scikit-learn) with A/B model versioning, reducing false-alert rate by 28% and operational overhead by 15% — cited in client delivery report.
Deployed inference models as Flask REST APIs serving 500+ predictions/day at sub-300ms p99 latency; designed API contract for future LLM reasoning layer integration.
Built Python automated feature engineering module ingesting 3 live data sources, eliminating 70% of manual prep and enabling continuous daily retraining cycles.

XGBoostFlask REST API28% Alert ReductionSub-300ms Latency

Jul 2024 — Aug 2024

Code Clause Pvt Ltd

Remote

Data Science Intern

Built NLP pipeline (tokenization, TF-IDF, cosine similarity) over 60K+ records; benchmarked SVD, KNN, MF, and Baseline collaborative-filtering models via RMSE and precision@K, lifting accuracy to 90%.

NLP Pipeline60K+ Records90% Accuracy

Selected Work

Projects

2026 · Featured

Constitutional AI-Aligned RAG Assistant with Full Evaluation Suite

Production RAG pipeline (embedding → FAISS → GPT-4 synthesis) with a Constitutional AI critique layer that self-checks responses against safety principles. Full RAGAS evaluation suite with reproducible benchmarks aligned with responsible AI deployment.

LangChainOpenAI APIFAISSRAGASConstitutional AIGPT-4

0.91

Ctx Precision

0.88

Faithfulness

0.93

Relevance

63%

Harm ↓

↗

2026 · Alignment Research

Transformer from Scratch with RLHF Reward Modeling

Decoder-only Transformer built entirely in PyTorch at 117M parameters, trained on WikiText-103. Layered an RLHF reward model using Bradley-Terry preference model + PPO fine-tuning — demonstrating the core alignment training loop used at frontier AI labs.

PyTorchHuggingFaceTRLRLHFPPOBradley-Terry

42.3

Perplexity

117M

Parameters

↗

2025 · Inference Engineering

GPU-Accelerated LLM Inference Engine

Optimized Mistral-7B for production via INT8 quantization, Flash Attention 2, and KV-cache management. Containerized with Docker, streaming FastAPI endpoint with sub-150ms TTFT at batch size 8.

PyTorchCUDATensorRTFastAPIDockerFlash Attention 2

3.2×

Throughput

<150ms

TTFT

↗

Under the Hood

Philosophy & Proficiency

shiva@ai-lab ~ bash

"The goal is not to build AI that is merely capable — but AI that is capable and trustworthy."

I believe the most important engineering work of our generation sits at the intersection of performance and alignment. Every system I build includes safety constraints, evaluation pipelines, and measurable benchmarks — as first principles, not afterthoughts.

LLM & RAG Engineering92%

PyTorch / ML Engineering88%

Alignment & RLHF85%

MLOps & Deployment80%

Academic Background

Education & Certifications

B.Tech, Computer Science & Engineering

SR University — Warangal, India

Aug 2021 — May 2025

CGPA 7.33 / 10

Deep LearningMachine Learning Data Structures & AlgorithmsLinear Algebra Probability & Statistics

🏆 1st Place — SR University Hackathon 2023 Grand Finale

Led 4-person team building a blockchain supply-chain transparency prototype. Outperformed 20+ teams, awarded for technical depth and live demo execution.

Certifications

◈

Oracle Cloud Infrastructure Generative AI Professional

Oracle · 2024

◈

Salesforce Certified AI Associate

Salesforce · 2024

◈

Advanced Data Analytics Specialization

Google

◈

Applied Data Science with Python Level 2

IBM

◈

MIT OpenCourseWare 6.034 (AI) Self-Study

HackerRank · LeetCode · Kaggle

Ready to Build

Let's Create
Something
Extraordinary

Open to full-time roles, research collaborations, and frontier AI projects.

✉ shivakumarvanam27@gmail.com ↓ Download Resume ↗ LinkedIn ↗ GitHub ↗ LeetCode

+91-9000773888 · Hyderabad, India

ShivaKumarVanam