Tejas K.

Key Skills

Software

No software listed

Top Subject Matter

LLM/RAG for semantic code search and MCP tooling

Local LLM inference integration for end-user AI assistant

ML model building and evaluation for time-series forecasting

Top Data Types

Text

Top Task Types

Fine-tuning

Freelancer Overview

Wordless — Local-First Semantic Codebase Search MCP Server. Brings 5+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include ChromaDB, Ollama, and scikit-learn. Education includes Bachelor of Technology, IIIT Raichur (2028). AI-training focus includes data types such as Computer Code and Programming and labeling workflows including Computer Programming, Coding, and Fine-tuning.

Labeling Experience

Gold Price Prediction — Time-Series Forecasting & Feature Engineering

Performed machine learning experimentation and evaluation for time-series regression to forecast gold prices using engineered features and multiple models. Conducted thorough EDA and stationarity/outlier analysis to validate feature relevance before training. Benchmarked models with MAE, RMSE, and directional accuracy to compare generalization performance. • Built regression pipelines using OHLC history and macroeconomic indicators • Engineered lag/rolling-window features for supervised learning • Evaluated Linear Regression, Random Forest, and XGBoost on held-out tests • Used MAE, RMSE, and directional accuracy for model selection

2024 - 2024

Mac AI Assistant — Naive macOS AI Productivity Tool (local LLM integration)

Fine-tuning

Developed an end-to-end LLM/RAG-capable local inference stack for a Rust-based macOS AI productivity assistant focused on on-device responses. Designed the application for low-latency, privacy-first usage on Apple Silicon by coupling the app with local model inference infrastructure. Supported product-level testing and iterative refinement from idea to release with tracked downloads. • Implemented local LLM inference integration for on-device responses • Built an end-to-end Rust macOS AI assistant application • Managed distribution, versioning, and user-facing polish independently • Delivered a user-facing tool emphasizing latency and privacy

2024 - 2024

Wordless — Local-First Semantic Codebase Search MCP Server

Built an LLM-powered local-first semantic code search MCP server using tree-sitter AST chunking and a vector database to retrieve relevant code. Integrated Qwen3 embeddings served via Ollama and used ChromaDB to store embeddings for search and retrieval. Evaluated retrieval quality against grep-based baselines to quantify improvements in efficiency and token usage. • Used Python with FastMCP to implement MCP server endpoints • Used ChromaDB vector storage and Ollama/llama.cpp for local embeddings • Implemented tree-sitter AST chunking to segment code into semantic units • Resolved transport and compatibility issues for Codex CLI integration

2024 - 2024

Education

Work History