Aditya K.

Key Skills

Software

Other

Top Subject Matter

Llms Domain Expertise

Rag Domain Expertise

AI search

Top Data Types

Text

Top Task Types

RLHF

Fine-tuning

Prompt + Response Writing (SFT)

Freelancer Overview

AI/ML Engineer (LLM and RAG Systems) - Perceptyx AI. Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other. Education includes Bachelor of Engineering, Visvesvaraya Technological University (2025). AI-training focus includes data types such as Computer Code, Programming, and Text and labeling workflows including RLHF, Fine-tuning, and Prompt + Response Writing (SFT).

Labeling Experience

AI/ML Engineer (LLM and RAG Systems) - Perceptyx AI

TextRLHF

Built an AI search engine that answers user queries using real-time web search with citations. Designed and implemented a multi-layer memory system to improve context retention and response quality across sessions. Developed intelligent query routing, search aggregation, and reranking to reduce latency through parallel retrieval and caching strategies. • Linked LLM and retrieval components with LangGraph and FastAPI for end-to-end query handling. • Implemented Redis semantic caching and async Qdrant queries to improve throughput and response times. • Built scalable background workflows with monitoring and RLHF-style self-improvement loops. • Integrated PostgreSQL for supporting data storage and operational persistence.

2025 - Present

Open-Source project — Secure RAG system with layered defenses (Aug 2025 - Present)

OtherPrompt + Response Writing (SFT)

Built a secure RAG system with multiple defenses to improve trustworthiness of retrieved and generated outputs. Implemented mechanisms to detect and mitigate prompt injection while filtering and separating role-based context. Focused on improving answer safety through layered input/output handling around retrieval and generation. • Added input sanitization to reduce malicious or malformed prompts. • Implemented prompt injection detection and output filtering for safer responses. • Enforced role-based context separation to prevent cross-role leakage. • Integrated these controls into a production-style RAG workflow.

2025 - Present

Open-Source project — Intelligent cloud anomaly detection system (Aug 2025 - Present)

OtherFine-tuning

Developed and applied LLM-assisted or ML-assisted components for intelligent cloud anomaly detection with reasoning to support model improvement. Focused on reducing false positives and improving incident response speed using hybrid ML plus reasoning approaches. Built evaluation-driven logic to refine detection decisions for operational deployment. • Implemented hybrid ML modeling together with LLM reasoning for anomaly classification. • Reduced false positives by tuning decision logic and response thresholds. • Accelerated incident response by improving detection confidence and actionability. • Built supporting pipelines for monitoring and optimization in production settings.

2025 - Present

Perceptyx AI — AI search engine with LangGraph/FastAPI/PostgreSQL and RLHF-based self-learning loops (Aug 2025 - Present)

OtherRLHF

Implemented RLHF-based self-learning loops to iteratively improve retrieval-augmented AI search behavior and response quality. Built evaluation/quality improvement workflows that support continuous optimization of model outputs over time. Engineered system components that enable feedback-driven refinement of search and generation results. • Reduced AI search latency using parallel retrieval, async vector search, and Redis semantic caching. • Implemented intelligent query routing, search aggregation, and reranking for better answer relevance. • Built a multi-layer memory (working, episodic, semantic, procedural) to retain context across sessions. • Developed scalable background workflows with monitoring for ongoing learning cycles.

2025 - Present

Open-Source AI Engineer (RAG, Memory, Anomaly Detection) - Open-Source

Text

Contributed to open-source AI projects focused on RAG systems, agent memory, and anomaly detection. Implemented security-focused safeguards such as prompt injection detection, input sanitization, and output filtering for safer retrieval. Developed intelligent anomaly detection by combining hybrid ML approaches with LLM reasoning to reduce false positives and speed up incident response. • Built secure RAG pipelines with multi-layer defenses and role-based context separation. • Designed AI memory modules to support long-term context retention in agentic multi-turn workflows. • Implemented hybrid ML and LLM reasoning techniques for anomaly detection and faster triage. • Worked across ML preprocessing, feature engineering, evaluation, optimization, and model fine-tuning.

2025 - 2025

Education

Work History