AI/ML engineer with 3+ years shipping enterprise GenAI — RAG architectures, Agentic AI and LLM orchestration — from raw data to a deployed, monitored endpoint. Transparent hourly rates, no agency markup.
Billed hourly, $5–9/hr depending on depth. No retainers locked behind contracts — start small, scale if it works.
Production GenAI, RAG and computer-vision work — open any card for the full challenge, approach and outcome.
An LLM-powered legal platform: lawyers upload case documents, run semantic legal search, track case workflows and get automated insights. RAG over Qdrant + Neo4j.
An Agentic AI layer over Repwr CRM: operate the CRM in natural language, retrieve insights and run autonomous deal & project workflows.
An AI CRM intelligence assistant: conversational querying, analytics retrieval and workflow automation via LLM reasoning and FastAPI microservices.
YOLOv8 road-damage detection (potholes, cracks, surface wear) from highway footage at 30+ FPS, deployed on AWS with geospatial tagging across 500+ km.
YOLOv5 defect detection on drone thermal & visual imagery of power-plant equipment — boilers, turbines, cooling towers — with automated reporting.
Every engagement runs the same transparent loop. You see progress weekly.
A short call to map the use-case, your data and the success metric. You get a fixed hour estimate.
Architecture for retrieval, models, vector store and guardrails — chosen for your latency and budget.
RAG or agents wired with FastAPI, evals and version control. Tracked, reproducible, reviewed weekly.
Deployed on AWS / Docker with monitoring, docs and a clean handover. You own the whole system.
I'm Sahil Chandel, an AI/ML engineer in New Delhi with 3+ years building production Generative AI — RAG systems, Agentic AI and LLM-powered platforms for legal, CRM and enterprise teams. I currently work as a Senior AI/ML Engineer and take on select freelance GenAI builds.
No account managers, no markup, no black boxes — you work directly with the engineer building your system, end to end: data, retrieval, models, API and deployment. You keep all of it.
The questions teams ask before we start. Anything else — just book a call.
Work is billed transparently at $5–9 per hour with a fixed hour estimate up front — no agency markup. A focused RAG assistant or chatbot usually lands in the low hundreds of dollars; a full agentic system with integrations and deployment costs more depending on scope.
RAG connects an LLM to your own documents and data so its answers are grounded in your knowledge instead of only its training. I build production RAG with hybrid search (BM25 + embeddings), a vector database like Qdrant and optional Neo4j knowledge graphs for accurate, source-backed responses.
Agentic AI doesn't just answer — it takes actions. Using LLM-driven intent detection and tool orchestration, an agent can create records, run workflows and call your APIs from a plain-language request. You need it when the assistant should operate your systems (CRM, internal tools), not just talk.
Both open and closed: Llama 3, Qwen 2.5, Mistral and GPT-OSS on the open side, plus Gemini, OpenAI and Claude. I pick and orchestrate models per task, latency and budget — and can keep everything self-hosted if your data must stay private.
A focused RAG assistant or chatbot is typically 1–3 weeks. A full agentic system with integrations, auth, evaluation and deployment usually runs 4–6 weeks, with weekly progress checkpoints.
Yes — that's a core part of my work. I've built agentic layers over CRM platforms (Repwr, Protly) that create deals, manage projects and retrieve data through conversation, all via secure FastAPI services with role-based access.
Yes. Systems are containerised with Docker and deployed on AWS (EC2, S3, Lambda) or fully on-prem with CI/CD, monitoring and audit logging — so sensitive data never has to leave your infrastructure.
Completely. You keep the code, the configs, the data and the deployment. No black boxes and no lock-in — you can run, extend or hand it to your own team at any time.
Tell me the use-case and the data you have. I'll reply within a day with an approach and an hour estimate.