AI/ML Engineer

Torry Harris Business Solutions; Bengaluru; On-site; INR 1,400,000 - 2,200,000; Full time

Job summary

Torry Harris Business Solutions in Bangalore is seeking an AI/ML Engineer with 2–3 years of hands-on experience to design, build, and ship AI features and agents into production. You will work on both LLM-based systems and traditional ML models, focusing on real-world performance, cost, and latency.

Required skills include Python with Flask, pandas, scikit-learn, LLM APIs, agent frameworks, and production RAG with vector databases.

Qualifications

• 2–3 years of experience in AI/ML with production-ready projects.
• Experience with LLMs and ML models in real-world deployments.
• Proficient in Python and ML libraries: pandas, scikit-learn.
• Experience with LLM APIs (Claude, OpenAI, Gemini) and tool-calling.
• Experience with agent frameworks (LangGraph, Claude Agent SDK, or OpenAI Agents SDK).
• Familiarity with production RAG, vector databases (pgvector or Qdrant).
• Experience with observability and monitoring of AI systems (Langfuse/LangSmith).
• Docker and Linux proficiency.
• Ability to explain and review generated code.

Responsibilities

• Ship AI features and agents end to end, from design to production.
• Write evals before shipping, then track quality, cost, and latency in production.
• Pick the right model for each task based on cost, latency, and accuracy.
• Protect against prompt injection and data leaks, and meet privacy and governance rules.

Skills

Python

Flask

Pandas

Scikit-learn

LLM APIs

LangGraph

Claude Agent SDK

OpenAI Agents SDK

MCP

RAG

Langfuse

LangSmith

Docker

Linux

Tools

pgvector

Qdrant

Kubernetes

GitLab CI

Job description

We are looking for an AI/ML Engineer with 2–3 years of experience, including at least 1 year on LLM or ML projects, to join our team in Bangalore. You will design, build, and ship AI features and agents into production, and you'll work on both LLM-based systems and classic machine learning models. This role suits someone who cares about how AI performs in the real world, not just in a demo, and who measures quality, cost, and latency as seriously as functionality.

Must have:

• Python with Flask , plus pandas and scikit-learn
• LLM APIs (Claude, OpenAI, Gemini): tool calling, structured outputs, streaming, and prompt caching
• Agent frameworks (LangGraph, Claude Agent SDK, or OpenAI Agents SDK), and MCP
• Production RAG: chunking, hybrid search, reranking, and vector databases (pgvector or Qdrant)
• Evals: test sets, LLM-as-judge, and regression checks before release
• LLM observability (Langfuse or LangSmith)
• Classic ML: classification, forecasting, and anomaly detection
• Docker and Linux
• Daily use of AI coding tools, with the ability to explain and review generated code

Good to have:

• Self-hosting models with vLLM or Ollama, including quantization
• Fine-tuning with LoRA or QLoRA
• Guardrails: prompt injection defense and PII redaction
• Multimodal models (vision, speech)
• Kubernetes, GitLab CI, Prometheus/Grafana

What we look for: Ownership, improving on the spec rather than just meeting it, raising risks early, and learning without being asked.

Roles and Responsibilities

• Ship AI features and agents end to end, from design to production
• Write evals before shipping, then track quality, cost, and latency in production
• Pick the right model for each task based on cost, latency, and accuracy
• Protect against prompt injection and data leaks, and meet privacy and governance rules