You will get RAG Support Assistant with Citations — Reduce Tickets, Faster Answers
Top Rated
Project details
Build a production-grade RAG assistant that answers with sources, stays fast, and deflects tickets. I’ll connect your docs/URLs (and optional read-only DB/API) to a retrieval-augmented pipeline so responses are grounded in your content—not guesswork. You get citations, guardrails, and a clean handover.
What’s included (all tiers)
• Working RAG with source citations
• Safety rules and basic evals
• Short runbook (update & test)
Starter (5 days) – Up to 500 items, hosted demo endpoint (no source code), 1 revision.
Standard (7 days) – Up to 2000 items, hosted embeddable widget (no source code), traces & latency metrics, docs, 2 revisions.
Advanced (10 days) – Hybrid retrieval (BM25 + vectors on Postgres/pgvector), expanded guardrails, staging/self-hosting optional with source code, 3 revisions.
Billing model after launch: support/updates via hourly (weekly payments) or monthly milestones; model usage is token-based (OpenAI/Azure).
What’s included (all tiers)
• Working RAG with source citations
• Safety rules and basic evals
• Short runbook (update & test)
Starter (5 days) – Up to 500 items, hosted demo endpoint (no source code), 1 revision.
Standard (7 days) – Up to 2000 items, hosted embeddable widget (no source code), traces & latency metrics, docs, 2 revisions.
Advanced (10 days) – Hybrid retrieval (BM25 + vectors on Postgres/pgvector), expanded guardrails, staging/self-hosting optional with source code, 3 revisions.
Billing model after launch: support/updates via hourly (weekly payments) or monthly milestones; model usage is token-based (OpenAI/Azure).
AI Algorithms
Large Language ModelAI Applications
AI Chatbot, AI Content Creation, AI Text-to-Image, AI Text-to-Speech, Automatic Speech Recognition, Conversational AI, Natural Language Generation, Natural Language Understanding, Synthetic Data GenerationAI Development Language
PythonAI Tools
GitHub CopilotAI Models
GPT-4What's included
| Service Tiers |
Starter
$299
|
Standard
$699
|
Advanced
$1,199
|
|---|---|---|---|
| Delivery Time | 5 days | 7 days | 10 days |
Number of Revisions | 1 | 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 | - | - |
Optional add-ons
You can add these on the next page.
Fast Delivery
+$150 - $250
Additional Revision
+$60Frequently asked questions
9 reviews
(7)
(2)
(0)
(0)
(0)
This project doesn't have any reviews.
ED
Erica D.
Jun 5, 2026
Operational intelligence platform seeking senior data engineer
Hernan was an absolute joy to work with. He is collaborative, kind, helpful and willing to step in and try to solve problems. He is trustworthy, does not burn through hours pretending to work if he's not and he is dependable and shows up. I enjoyed working with him on my project and he helped us advance far on the project. I highly recommend him if you're looking for someone you can trust.
TG
Tyler G.
Apr 24, 2026
Equivalent to 8 hour contract - RAG Pipeline/Conversation Optimization
HL
Horacio L.
Dec 30, 2025
Database Administrator/Data Engineer with Retool Experience
Hernan is an exceptional developer and an absolute pleasure to work with. From day one, he was thoughtful, communicative, and genuinely invested in understanding our business needs—not just completing tasks.
The quality of his work was excellent. He approached complex problems with clarity, offered smart suggestions, and consistently delivered clean, well-structured solutions. Even when requirements evolved, Hernan remained flexible, proactive, and solution-oriented.
Although we ultimately decided to pause and shift away from a custom-built solution for internal strategic reasons, this decision had nothing to do with Hernan’s performance. His work exceeded our expectations, and we would not hesitate to work with him again in the future.
If you’re looking for a reliable, skilled, and professional developer who truly cares about doing great work, I highly recommend Hernan. We look forward to collaborating with him again when the opportunity arises.
The quality of his work was excellent. He approached complex problems with clarity, offered smart suggestions, and consistently delivered clean, well-structured solutions. Even when requirements evolved, Hernan remained flexible, proactive, and solution-oriented.
Although we ultimately decided to pause and shift away from a custom-built solution for internal strategic reasons, this decision had nothing to do with Hernan’s performance. His work exceeded our expectations, and we would not hesitate to work with him again in the future.
If you’re looking for a reliable, skilled, and professional developer who truly cares about doing great work, I highly recommend Hernan. We look forward to collaborating with him again when the opportunity arises.
HL
Horacio L.
Dec 15, 2025
Database Administrator/Data Engineer with Retool Experience
Hernan was one of the best freelancers we have ever worked with. From day one, he demonstrated exceptional efficiency, deep technical knowledge, and an impressive ability to understand our business needs quickly. His work on our database build was fast, accurate, and thoughtfully executed.
He is extremely open-minded, receptive to feedback, and collaborative in finding the best solutions. Hernan consistently went above and beyond, and it’s clear he’s one of the top professionals in his field.
We would highly recommend Hernan to anyone looking for someone reliable, skilled, and truly excellent at what they do.
He is extremely open-minded, receptive to feedback, and collaborative in finding the best solutions. Hernan consistently went above and beyond, and it’s clear he’s one of the top professionals in his field.
We would highly recommend Hernan to anyone looking for someone reliable, skilled, and truly excellent at what they do.
TG
Tyler G.
Dec 6, 2025
Improvement of Existing System
About Hernan
Senior Backend & AI Systems Architect, RAG, Vector Search, AI Agents
100%
Job Success
Punta del Este, Uruguay - 4:12 pm local time
I build production‑grade AI features and high‑performance backends. Retrieval‑Augmented Generation (RAG), agent workflows, vector search, text‑to‑SQL, and evaluation pipelines — glued to solid engineering. I have 15+ years shipping scalable systems and ~2 years building applied GenAI for a startup and client projects.
I design “AI‑first” systems where LLMs are controlled, observable, and secure: a staged pipeline (Planner → Clarifier → SQL‑Gate → Executor → Synthesizer) with checkpoints, read‑only DB access, output validation, and clear I/O contracts. This yields answers you can trust, trace, and scale.
Core AI Skill Set
• RAG systems: retrieval pipelines, chunking, hybrid search (lexical + dense), reranking, grounding & citations.
• Vector search on Postgres with pgvector: schema design, ANN indexes (HNSW/IVFFlat), recall/latency tuning.
• Agentic orchestration (LangGraph‑inspired): stateful graphs, checkpoints/resume, human‑in‑the‑loop interrupts, streaming, and debugging.
• Text‑to‑SQL: constrained decoding (e.g., PICARD‑style prompts) + SQL AST validation; parameterized queries only.
• LLM safety & governance: OWASP Top‑10 controls (prompt‑injection mitigations, secure tool use, output validation, least‑privilege data access), secrets handling, rate limiting, and audit trails.
• Observability & evals: step‑level logs/traces, latency and answer‑quality metrics, fixtures from real traces for regression.
Services & Packages
• RAG MVP (2–4 weeks): data connectors, normalization + chunking/embeddings, hybrid retrieval, grounding/citations, basic evals, and a clean chat UI.
• Agentic Analytics: Planner/Clarifier/SQL‑Gate/Executor pipeline for BI over your data; JSON‑first responses, guardrails, dashboards for query quality.
• LLM Safety Review: gap assessment vs OWASP LLM Top‑10 + practical remediation roadmap and hardening checklist.
• Postgres Vectorization: pgvector rollout, ANN index tuning (HNSW/IVFFlat), latency/recall SLAs, and hybrid retrieval with BM25/trigram.
Representative AI Projects
• E‑commerce RAG Assistant: multi‑tenant product QA with hybrid retrieval, ANN search, and guardrails; human approval for sensitive SQL; measurable lift in accuracy and latency.
• Agent Workflow for Analytics: Planner→Clarifier→SQL‑Gate→Executor pipeline; constrained decoding ensures valid SQL across unseen schemas; reduced query failures and hallucinations.
• Safety Hardening for LLM Apps: implemented OWASP controls (input filtering, tool allow‑lists, output validation, secrets isolation); added step‑level telemetry for audits.
Backend & Systems (Non‑AI)
• Scalable backend development: PHP (Laravel/CodeIgniter), Python, Node.js.
• Distributed systems & performance: caching, queues, load balancing, high‑concurrency patterns.
• Database architecture & optimization: Postgres/MySQL, Redis/Memcached; schema design, indexing, query tuning.
• APIs & integrations: secure REST/JSON APIs, third‑party integrations, webhooks, auth and permissions.
• Cloud & DevOps: Linux, AWS, Serverless, CI/CD, infrastructure as code, observability.
How I Work
• Business‑first mindset, clear communication, and fast iteration.
• Clean, maintainable, and testable code; contracts between stages (JSON schemas) for reliability.
• Performance and reliability at scale; secure defaults (read‑only DB access, timeouts, rate limits).
Let’s build AI that ships — reliable, secure, and fast. Click “Message” and tell me about your project.
Steps for completing your project
After purchasing the project, send requirements so Hernan can start the project.
Delivery time starts when Hernan receives requirements from you.
Hernan works on your project following the steps below.
Revisions may occur after the delivery date.
Kickoff & asset check
We review the use case, success metric, and knowledge sources. I confirm scope, risks, and delivery timeline based on your materials.
Data ingest & indexing
I ingest provided URLs/files, normalize content, and build the retrieval index (BM25 or hybrid depending on tier). Quick sanity tests to ensure coverage.