Open to Work·Bengaluru · 2026
Rakshith Kumar K.N
Senior AI/ML Engineer · 5+ yrs · GenAI, RLHF & Multi-Agent Systems
LangGraphGraphRAGPrompt EngineeringAWS BedrockRLHF / DPOMCPDeepEvalDSPyHIPAA
5+
Years Exp.
$2M+
Cost Saved
10K+
Daily AI Queries
99.99%
Uptime SLA
Senior AI/ML Engineer
Rakshith Kumar K.N
# rakshith.py
class RakshithKumar:
role = "Senior AI/ML Engineer"
stack = ["LangGraph", "RLHF", "DSPy"]
yoe = 5
def ship(self, idea):
return "production-ready AI"
GraphRAG ·
AWS Bedrock ·
LangGraph ·
HIPAA Infra ·
RLHF / DPO ·
vLLM ·
Neo4j ·
PEFT / LoRA ·
Kubernetes ·
Azure AI Foundry ·
TensorRT ·
MCP / A2A Agents ·
DeepEval / Ragas ·
Promptfoo ·
DSPy ·
Langfuse ·
TensorFlow ·
OpenAI Agents SDK ·
NeMo Guardrails ·
Alignment Engineering ·
CI/CD ·
GitHub Actions ·
GraphRAG ·
AWS Bedrock ·
LangGraph ·
HIPAA Infra ·
RLHF / DPO ·
vLLM ·
Neo4j ·
PEFT / LoRA ·
Kubernetes ·
Azure AI Foundry ·
TensorRT ·
MCP / A2A Agents ·
DeepEval / Ragas ·
Promptfoo ·
DSPy ·
Langfuse ·
TensorFlow ·
OpenAI Agents SDK ·
NeMo Guardrails ·
Alignment Engineering ·
CI/CD ·
GitHub Actions ·
01 · About
AI Engineer.
Infrastructure Thinker.
Available
I'm Rakshith Kumar K.N — a Senior AI/ML Engineer with 5+ years delivering $2.3M+ measurable business impact across enterprise security, healthcare, and fintech. MCP Foundation Development contributor building production MCP servers connecting LLM agents to live enterprise tools.
Proven team leader who mentored 5+ junior engineers, led cross-functional squads of 4–8 engineers, and pioneered LLM-supervised Knowledge Distillation — compressing 4+ hour firewall validation to 2–5 seconds. Multi-cloud architect across AWS, Azure, and GCP with $50K+/month in cloud savings.
M.Sc. Big Data Analytics from St. Joseph's University, Bengaluru. B.C.A. from Seshadripuram College. Currently Senior AI/ML Engineer at Gruve AI building enterprise-grade systems for Fortune 500 clients including Cisco and J.P. Morgan.
Deep expertise in multi-agent orchestration, GraphRAG, LLMOps pipelines, and regulated-domain AI (HIPAA, SOC 2). Zero PII leakage across 50K+ financial documents.
Education
M.Sc. Big Data Analytics
St. Joseph's University, Bengaluru · 2020–2022
B.C.A. Computer Applications
Seshadripuram College, Bengaluru · 2016–2019
Awards & Recognition
★
Winner — Gruve AI Hackathon
IT Help Desk automation project
★
Winner — August AI Hackathon
AI Doctor Analyzer for clinical records
Co-author — IEEE ACAI 2022
Water Quality Prediction using BPNN, SVR, and LSTM
Published at IEEE International Conference on Advanced Computing and Artificial Intelligence
★
MCP Foundation Development Contributor
Open-source Model Context Protocol ecosystem
LLMs & GenAI
GPT-4 / Claude 3.5LangGraphGraphRAGMCP ProtocolRAGPEFT/LoRAKnowledge DistillationFine-tuning
Cloud & Infra
AWS BedrockAzure AI FoundryGCP Vertex AITerraformKubernetes
ML / DL / Inference
PyTorchCUDA / TensorRTvLLM / SGLangSDXLBERT / YOLOControlNet
// profile.json
{
"name": "Rakshith Kumar K.N",
"title": "Senior AI/ML Engineer",
"location": "Bengaluru, India",
"experience": 5,
"education": ["MSc Big Data Analytics", "BCA"],
"current": "Gruve AI",
"clients": ["Cisco", "J.P. Morgan"],
"speciality": [
"Generative AI",
"Multi-Agent Systems",
"HIPAA Infra",
"Diffusion Inference"
],
"open_to_work": true
}
02 · Experience
Where I've Shipped.
Metrics
Clients
Tech
Leadership
Oct 2024 – Present
Gruve AI
CurrentSenior AI/ML Engineer
→Custom Eval Framework + Knowledge Distillation (Cisco ASA) — GPT-4-supervised eval pipeline with KL-divergence loss and confidence-threshold gating; trained 3 student models achieving zero LLM calls at inference; 98% reduction (4+ hours → 2–5 sec)
98% reduction200+ firewallsCisco ASAEval harnessGPT-4Led 4 engineers
→Production MCP Server Development — MCP servers connecting LLM agents to HappyFox, Jira, Confluence for real-time enterprise tool integration; applied DSPy for prompt optimization
85% latency cutMCP SDKDSPyLed 3-person squad
→IDP Supply Chain Agent (Scania) — 5-stage confidence-gated agent harness for shipping document intelligence; hybrid NL-to-SQL + semantic search; custom eval framework with per-field precision thresholds as LLM-as-judge layer
p95 <300ms73% HITL reductionZero critical errorsScaniaAWS BedrockElasticsearch
→TechGenie Voice Agent [MCP-Powered] — MCP-connected LLM agents with NeMo Guardrails for safe output control; DSPy optimization reducing hallucination by 34%
42% faster resolution34% hallucination ↓500+ weeklyNeMo GuardrailsDSPyLed 4-engineer team
→Enterprise AI Search (GSC) — Hybrid graph + vector retrieval (1M+ graph nodes, 50M embeddings) at 99.9% uptime
10K+ queries/dayp95 <300ms99.9% uptimeGraphRAGNeo4jMentored 2 engineers
→AI Safety - PII Redaction (J.P. Morgan) — NLP redaction pipeline achieving zero PII leakage across 50K+ financial documents; scaled to 2,000+ docs/day
Zero PII leakage50K+ documents99.2% precisionJ.P. MorganSOC 2 passed
→GUIS Meeting Intelligence Platform — 6-service platform for real-time meeting insights across Google Meet & MS Teams; 45% reduction in cross-team meeting follow-up time
45% follow-up ↓200+ weekly meetings6 microservicesGeminiDeepgramLed 5 engineers
→Multi-Cloud FinOps — Migrated VMs to event-driven containers across AWS, Azure, GCP with CI/CD pipelines; 99.5% SLA
$18K/month saved62% compute reduction99.5% SLATerraformGitHub Actions
→Team Leadership & Mentoring — Mentored 5 junior ML engineers (2 promoted within 8 months); established org-wide documentation standards and technical design reviews
Mentored 5 engineers2 promoted in 8 mo
Aug 2023 – Oct 2024
August AI
AI Engineer
→Multi-Agent Patient Interview System — Context-aware multi-agent prompt pipeline raising diagnostic accuracy 67% → 91%; 2,500+ daily interactions across 3 hospitals
67% → 91% accuracy2,500+ daily3 hospitalsLangGraphDSPyLed 3-person ML team
→LLM Evaluation and Safety Pipeline — RAGAS evaluation harness measuring faithfulness, relevance, and context precision; confidence-threshold HITL gating; eval-driven DSPy optimization loop
5,000+ sessions evaluatedRAGASDSPyLangSmith
→RLHF & Alignment Engineering (Qwen 2.5 72B) — Full alignment pipeline: LoRA/QLoRA instruction tuning → DPO on clinician-preference pairs; trained reward model on domain-specific human preference data
Qwen 2.5 72BDPO / RLHFLoRA/QLoRAUnslothModal + RunPod
→HIPAA-Compliant Infrastructure — 99.99% uptime for 500+ users across Azure and AWS; CI/CD via GitHub Actions; passed 2 regulatory audits with zero findings
99.99% uptime2 audits passedHIPAAGitHub ActionsMLflow
→Clinical NLP Pipeline — Automated document review with 97% accuracy; 85% manual review reduction
$150K annual savings97% accuracy10K+ docs/monthTrained 2 analysts
Jan 2023 – Aug 2023
Krut AI (Snive)
Machine Learning Engineer
→Generative Image Pipeline — TensorRT + Fast UNet + VAE optimizations on AWS G5 instances; SDXL + ControlNet for e-commerce at scale
8s → 1.2s (85%↓)50K+ images/moTensorRTControlNetMentored 2 engineers
→GPU Cost Optimization & MLOps Pipeline — Mixed precision + gradient checkpointing (SSIM >0.95); automated model versioning and deployment with MLflow + GitHub Actions + ONNX; deployment from 3 days → 4 hours
$32K/month saved3 days → 4hr deploySSIM >0.95MLflowGitHub ActionsONNX
Mar 2022 – Jan 2023
Lincode Labs Inc.
Associate Data Scientist
→Event-Driven Architecture — Replaced REST polling with SSE for real-time inference; $2M/year savings across 15 global facilities
$2M/year savings450M req/month15 global facilitiesSSE
→CV/OCR Model Delivery & Monitoring — ONNX runtime optimization for 8 international clients; Prometheus monitoring at 450M req/month; reduced support tickets 60% by training 20+ client engineers
92% accuracy0.24s inference60% ticket ↓8 clientsTensorFlowTrained 20+ engineers
03 · Skills
Skill Levels.
Recruiter-scannable proficiency — at a glance
GenAI & LLMs
Prompt Engineering93%
GPT-4 / Claude 3.5 / Gemini Pro95%
LangGraph92%
GraphRAG / RAG92%
Multi-Agent Systems90%
MCP Protocol88%
RLHF / DPO / PPO85%
Alignment Engineering84%
LLM-as-Judge / Confidence-Gated Eval82%
PEFT / LoRA / QLoRA85%
Knowledge Distillation87%
Guardrails / Structured Outputs80%
LLM Red-Teaming (Promptfoo)78%
Fine-tuning84%
Cloud & MLOps
AWS (Bedrock, SageMaker, ECS, Lambda)90%
Azure (Container Apps, Cognitive Svcs)85%
GCP (Vertex AI, Cloud Run, Gemini)82%
Terraform / IaC80%
Kubernetes / Docker83%
MLflow / LangSmith86%
Langfuse82%
Arize Phoenix78%
LLM Evaluation (DeepEval / Ragas)80%
CI/CD (GitHub Actions)82%
Prometheus / Monitoring80%
ML / Deep Learning
PyTorch88%
TensorFlow80%
CUDA / TensorRT82%
vLLM / TensorRT-LLM / SGLang80%
SDXL / Diffusion78%
BERT / YOLO85%
Whisper / Speech78%
ONNX / Quantization82%
Data & Vector
Neo4j85%
Qdrant83%
Pinecone78%
PostgreSQL / MongoDB82%
Elasticsearch80%
Redis81%
MinIO75%
AI Frameworks
LangChain90%
LlamaIndex82%
DSPy80%
OpenAI Agents SDK82%
Google ADK78%
NeMo Guardrails78%
AutoGen / CrewAI80%
Semantic Kernel75%
Hugging Face86%
Unsloth / LLaMA Factory78%
04 · Projects
What I've Built.
↗
GraphRAG · Enterprise Search
GSC — AI-Powered Search Engine
1M+ graph nodes, 50M embeddings, hybrid graph + vector retrieval at p95 <300ms with 99.9% uptime.
<300ms retrieval1M+ nodes99.9% uptime
Neo4jQdrantAWS BedrockECS
↗
Agent Harness · Supply Chain
IDP Supply Chain Agent (Scania)
5-stage confidence-gated agent harness for shipping document intelligence; 73% HITL reduction with zero critical-field errors.
73% HITL ↓p95 <300msZero critical errors
AWS BedrockClaudeElasticsearch
↗
Healthcare AI · Multi-Agent
AI Doctor Analyzer
Multi-agent patient interview system raising diagnostic accuracy 67% → 91%. August AI Hackathon winner. 2,500+ daily interactions across 3 hospitals.
91% accuracy2,500+ daily3 hospitals
LangGraphGPT-4DSPy
↗
Alignment · RLHF
RLHF & Alignment Pipeline (Qwen 72B)
Full alignment pipeline for Qwen 2.5 72B medical LLM — LoRA instruction tuning + DPO on clinician-preference pairs with trained reward model.
72B paramsDPO / RLHFReward Model
PyTorchUnslothLLaMA Factory
↗
Voice Agent · IT Support
Gruve TechGenie
MCP-powered voice IT triage with NeMo Guardrails for safe output control; DSPy optimization reducing hallucination by 34%. 500+ weekly interactions.
↓42% resolution↓34% hallucination
NeMo GuardrailsDSPyMCP
↗
Diffusion · Inference Opt.
Fast SDXL Pipeline
Fast UNet + VAE for SDXL + ControlNet — 8s → 1.2s inference. $32K/month GPU savings. Automated MLOps deployment pipeline.
1.2s inference↓85% latency$32K/mo saved
SDXLTensorRTGitHub Actions
↗
NLP · Compliance
J.P. Morgan Redaction Service
Enterprise PII redaction (spaCy + Presidio + GLiNER) for 50K+ financial documents. Zero PII leakage. 2,000+ docs/day. SOC 2 passed.
Zero PII leakage50K+ docsSOC 2
spaCyPresidioGLiNER
↗
Knowledge Distillation · Security
Cisco Firewall Validation (FMT v2.0)
GPT-4-supervised eval + distillation pipeline — custom eval framework with KL-divergence loss and confidence-threshold gating. 4+ hours → 2–5 sec.
2–5s validation98% reduction200+ firewalls
GPT-4PyTorchBERT
↗
Computer Vision · Manufacturing
Industrial OCR & Detection
OCR + detection for 8 manufacturing clients. Prometheus monitoring at 450M req/month. 60% support ticket reduction. TensorFlow + ONNX.
92% accuracy$2M/yr60% ticket ↓
PyTorchTensorFlowONNX
↗
Infrastructure · ML Serving
Distributed Inference Service
Multi-region ML serving at 5K req/sec with 99.9% uptime across AWS and GCP. Open-source.
5K req/sec99.9% uptime
Ray ServeRedisTerraform
↗
Product · Meeting Intelligence
GUIS Meeting Intelligence Platform
6-service platform for real-time meeting insights across Google Meet & MS Teams. 45% reduction in follow-up time across 200+ weekly meetings.
45% follow-up ↓6 microservices200+ weekly
ReactFastAPIGeminiDeepgram
↗
Open Source · MCP Protocol
MCP Server Development
Production MCP servers connecting LLM agents to HappyFox, Jira, Confluence. 85% latency reduction. DSPy-optimized. Foundation Team contributor.
85% latency cutFoundation Team
MCP SDKTypeScriptGoDSPy
05 · AI Interface
Ask Me Anything.
Powered by AI · Trained on Rakshith's resume · Can send him an email directly.
RAKSHITH.AI
AI-powered · Context-aware · Markdown enabled
06 \u00B7 Contact
Let's Build Together.
↓ Download ResumeCurrently open to Staff / Senior AI Engineer roles at FAANG, AI Labs, and high-growth startups. Excited about GenAI infrastructure, agentic systems, healthcare AI, and enterprise RAG.
Based in Bengaluru, India. Open to remote, hybrid, or relocation. Response: under 24 hours.
Let's start something
great together.
great together.