You will get an LLM evaluation harness and a report on exactly where your model fails
Top Rated
Top Rated
Project details
Most teams shipping LLM features can't answer one question: is the new prompt or model actually better, or does it just feel better? Without an eval set, six-figure decisions get made on vibes.
I build the measurement layer first. You get a failure-mode breakdown, where your model hallucinates, breaks format, degrades on long context, or over-refuses, with the exact inputs that trigger each one, instead of a single score that hides it all. Then a harness you own outright, runnable in your CI as a regression gate, so what you fix stays fixed.
What sets this apart: almost nobody sells the measurement. I calibrate the LLM-as-judge against human labels and report the agreement (Cohen's κ) rather than asking you to trust it, and I flag the cheapest model that clears your bar, not the highest-scoring one. Roughly a third of the time the audit shows the real problem is retrieval or prompting and a fine-tune wouldn't have helped — you'll hear that before you fund it, not after.
I build the measurement layer first. You get a failure-mode breakdown, where your model hallucinates, breaks format, degrades on long context, or over-refuses, with the exact inputs that trigger each one, instead of a single score that hides it all. Then a harness you own outright, runnable in your CI as a regression gate, so what you fix stays fixed.
What sets this apart: almost nobody sells the measurement. I calibrate the LLM-as-judge against human labels and report the agreement (Cohen's κ) rather than asking you to trust it, and I flag the cheapest model that clears your bar, not the highest-scoring one. Roughly a third of the time the audit shows the real problem is retrieval or prompting and a fine-tune wouldn't have helped — you'll hear that before you fund it, not after.
AI Algorithms
Large Language Model, Multilayer Perceptron, Recurrent Neural NetworkAI Applications
AI Content Creation, Conversational AI, Natural Language Generation, Natural Language UnderstandingAI Development Language
PythonAI Tools
Hugging Face, PyTorchAI Models
LLaMAWhat's included
| Service Tiers |
Starter
$1,200
|
Standard
$2,900
|
Advanced
$5,500
|
|---|---|---|---|
| Delivery Time | 5 days | 10 days | 18 days |
Number of Revisions | 2 | 2 | 2 |
AI Model Integration | - | ||
Batch Normalization | - | - | - |
Database Integration | - | - | |
Detailed Code Comments | - | ||
Image Upscaling | - | - | - |
MLOps | - | ||
Model Deployment | - | - | - |
Model Documentation | - | ||
Model Monitoring | - | - | |
Model Testing & Optimization | |||
Model Tuning | - | - | - |
Natural Language Processing | |||
NLP Tokenization | - | - | - |
Pre-Training | - | - | - |
Prompt Engineering | |||
Setup File | - | ||
Source Code | - |
Frequently asked questions
26 reviews
(26)
(0)
(0)
(0)
(0)
This project doesn't have any reviews.
AV
Ana V.
Jun 19, 2025
Machine Learning Engineer
I’m thrilled to share my experience working with Shiva on our recent LLM project:
Shiva’s depth of knowledge in large language models is very impressive. From day one, he demonstrated a clear grasp of cutting-edge architectures, prompt engineering best practices, and fine-tuning strategies that perfectly aligned with our goals.
What sets Shiva apart is his razor-sharp attention to detail. Whether it was debugging unexpected tokenization issues or calibrating our evaluation metrics, he spotted subtle inconsistencies before they grew into larger obstacles and fixed them.
He’s also super responsible. Deadlines were never missed, dependencies were always tracked, and deliverables arrived on time, fully documented, and ready for immediate use.
Beyond technical skill and reliability, Shiva brings genuine curiosity and problem solving drive to every issue we face. He doesn’t just apply known solutions, he asks “Why?” and “What if…?”, which led us to innovative tweaks that improved our model’s accuracy by over 15%.
In short, Shiva is the LLM expert you want on your team: knowledgeable, dependable, detail-oriented, and passionately committed to solving tough problems. I couldn’t recommend him more highly.
Shiva’s depth of knowledge in large language models is very impressive. From day one, he demonstrated a clear grasp of cutting-edge architectures, prompt engineering best practices, and fine-tuning strategies that perfectly aligned with our goals.
What sets Shiva apart is his razor-sharp attention to detail. Whether it was debugging unexpected tokenization issues or calibrating our evaluation metrics, he spotted subtle inconsistencies before they grew into larger obstacles and fixed them.
He’s also super responsible. Deadlines were never missed, dependencies were always tracked, and deliverables arrived on time, fully documented, and ready for immediate use.
Beyond technical skill and reliability, Shiva brings genuine curiosity and problem solving drive to every issue we face. He doesn’t just apply known solutions, he asks “Why?” and “What if…?”, which led us to innovative tweaks that improved our model’s accuracy by over 15%.
In short, Shiva is the LLM expert you want on your team: knowledgeable, dependable, detail-oriented, and passionately committed to solving tough problems. I couldn’t recommend him more highly.
HL
Hui Sze L.
Jun 19, 2025
Pytorch Tutor
AP
Arun P.
Apr 29, 2025
30 minute consultation
AV
Ana V.
Apr 29, 2025
30 minute consultation
Very knowledgable
FS
Farzad S.
Feb 27, 2025
3h LLM Architecture Consultation Work
Shivanand did an outstanding job helping me conceptualize a software architecture for an agentic AI system. His deep knowledge of LLMs, NLP, and AI frameworks was truly impressive. He not only provided valuable insights but also communicated complex ideas in a clear and structured manner. His expertise and professionalism made the collaboration seamless and highly productive. I highly recommend Shivanand to anyone looking for a top-tier AI expert!
About SHIVANAND
Senior LLM Engineer | Fine-Tuning, RAG, Agents & Self-Hosted Serving
100%
Job Success
Bengaluru, India - 8:33 am local time
Five years training language models, and five years shipping systems built on them. That combination is why teams call me when what they already built stops holding up.
WHAT I DO
1) Production LLM systems - RAG, agents, serving
Retrieval that actually retrieves: 92% retrieval accuracy on a LlamaIndex + Weaviate pipeline with a 20% cut in query time. Chunking, embedding choice, hybrid and reranked retrieval - and an eval set that proves the change helped instead of a vibe check.
Agent systems with a cost and latency budget: multi-agent pipelines in LangChain / LlamaIndex / CrewAI, tool calling, long-running state. One automated product-information system cut manual review effort ~90%.
Self-hosted and open-weight serving: GPU sizing, quantisation strategy, throughput and concurrency planning, and quality validation against a frontier baseline before you cut over. At Dell I shipped 4-bit GPTQ quantisation and SparseGPT pruning (~40% sparsity) for hardware-constrained inference.
Latency and cost: replacing an LLM call with a fine-tuned 300M classifier took one production path from ~1s to ~100ms. Model routing, caching, honest per-request cost accounting.
Evaluation and regression gates: offline eval sets, LLM-judge calibration, CI gates so a prompt or model change can't silently regress. Most teams I meet have no way to answer "is it better than last week."
2) Fine-tuning, post-training and alignment
Pre-trained a 355M-parameter GPT-2-medium architecture from scratch on 28B tokens (Cosmopedia-v2), distributed across 4x NVIDIA H100's with DeepSpeed - mixed precision, gradient accumulation, LR scheduling. Beat the original GPT-2-medium checkpoint on perplexity.
Improved Phi-4-14B-Instruct by 2% across every Hugging Face leaderboard benchmark via Model Stock merging, validated cheaply first on a LoRA-tuned Qwen2.5-1.5B proxy over 1.2M curated STEM samples.
LoRA-tuned Qwen2.5-14B-Instruct on 12K reasoning samples, using synthetic data from a multi-agent generation pipeline - measured gains on GSM8K, GPQA, and MMLU.
LoRA + DPO on Llama-2 for customer-care summarisation (23K SFT samples, 5K preference pairs): 17% better across evaluation metrics.
Designed and ablated a novel Drift-Diffusion attention mechanism on BERT-base, with full Weights & Biases tracking across baseline, unscaled and gated variants.
The honest version: most projects that arrive asking for a fine-tune don't need one. The base model was already good enough, the eval set couldn't detect improvement, or the problem was retrieval. I'll tell you which before you spend GPU budget - that answer is worth more than the training run.
STACK
Python, PyTorch, Hugging Face, DeepSpeed, Weights & Biases, FastAPI, LangChain, LlamaIndex, CrewAI, Weaviate, Elasticsearch, MongoDB, Docker, Kubernetes, AWS (EC2, Inferentia-2), GCP. OpenAI, Anthropic, Gemma / Llama / Mistral / Qwen / Phi. ElevenLabs and LiveKit for voice.
BACKGROUND
ML Engineer at Dell Technologies and BYJU'S AI Labs, where a multi-objective Transformer recommender I built served 1M+ students across 1B+ data points at 85% F1. Contributor to Hugging Face Transformers documentation and to DocsGPT. Top-Rated on Upwork with a 100% Job Success Score.
WHO I'M NOT FOR
If the job is wiring Zapier or n8n between two SaaS tools, hire a generalist - genuinely, you'll get a better deal and a faster one. I'm worth the rate when the system has to be correct, cheap and measurable under real traffic.
HOW TO START
Start with the fixed-fee diagnostic rather than an open-ended hourly build. I read your pipeline, your traces and your evaluation setup, then send a written diagnosis: where quality is leaking, what each request actually costs, what to fix first, and what fixing it takes. It stands on its own as a deliverable, and it becomes the scope if you want me to do the build.
Send me what's breaking and one example of the wrong output. That's enough to start.
Steps for completing your project
After purchasing the project, send requirements so SHIVANAND can start the project.
Delivery time starts when SHIVANAND receives requirements from you.
SHIVANAND works on your project following the steps below.
Revisions may occur after the delivery date.
End to end steps are below
- Scoping call — define what "wrong" means - Build the eval set from your traffic - Benchmark your model against alternatives - Build the failure taxonomy - Calibrate the judge and wire up the harness (Standard and Advanced) - Handover walkthrough