Building a Smart Placement Mentor Agent with Amazon Bedrock, Strands SDK, and MCP

from strands import Agent
from strands.models.ollama import OllamaModel

ollama_model = OllamaModel(
host="http://localhost:11434",
model_id="llama3.2:3b"
)

agent = Agent(
model=ollama_model,
tools=[],
system_prompt="""You are an AI Placement Assistant.
Company eligibility rules:
- TCS: CGPA >= 6.0, no backlogs
- Infosys: CGPA >= 6.5, no backlogs
- Amazon: CGPA >= 8.0, CS/IT degree, no backlogs
..."""
)

from strands import Agent, tool
from strands.models.bedrock import BedrockModel

@tool
def eligibility_checker_tool(cgpa: float, backlogs: int, degree: str) -> str:
"""
Check which companies a student is eligible for.

Args:
cgpa: Student CGPA
backlogs: Number of active backlogs
degree: Student degree (e.g. B.Tech, MCA)

Returns:
List of eligible and not-eligible companies with reasons
"""
companies = {
"TCS": {"min_cgpa": 6.0, "max_backlogs": 0},
"Infosys":{"min_cgpa": 6.5, "max_backlogs": 0},
"Amazon": {"min_cgpa": 8.0, "max_backlogs": 0},
# ...
}
# check logic here

from strands_tools import mem0_memory

agent = Agent(
model=bedrock_model,
tools=[mem0_memory],
system_prompt="""You are an AI Placement Assistant with persistent memory.

MEMORY RULES:
- When a student shares their profile (name, CGPA, branch, skills),
ALWAYS store it immediately using your memory tool.
- At the start of every session, recall memory to check if a profile exists.
- Always greet the student by name if you remember them.
"""
)

def stream_callback(**kwargs):
if "data" in kwargs:
print(kwargs["data"], end="", flush=True)
elif "current_tool_use" in kwargs and kwargs["current_tool_use"].get("name"):
print(f"\n🔧 Using tool: {kwargs['current_tool_use']['name']}", flush=True)

agent = Agent(
model=bedrock_model,
tools=[...],
callback_handler=stream_callback
)

You: What should I learn to get into Amazon?

🔧 Using tool: recall_memory
🔧 Using tool: skill_gap_analyzer_tool

Agent: Hi Kaniska! Based on your profile (CGPA 8.8, Python + SQL):

✅ You have: Python, SQL
❌ Missing: DSA (LeetCode level), System Design, Java or C++

📚 Learning Roadmap:
1. DSA — LeetCode Easy → Medium → Hard (start with Arrays, HashMap)
2. System Design — Gaurav Sen YouTube playlist
3. Java or C++ — pick one, complete a full course in 4–6 weeks

from strands.tools.mcp import MCPClient
from mcp import StdioServerParameters, stdio_client

mcp_client = MCPClient(
lambda: stdio_client(
StdioServerParameters(
command="uvx",
args=["awslabs.aws-documentation-mcp-server@latest"],
env={"FASTMCP_LOG_LEVEL": "ERROR", **os.environ}
)
)
)

with mcp_client:
mcp_tools = mcp_client.list_tools_sync()

agent = Agent(
model=bedrock_model,
tools=[
*local_placement_tools, # all tools from Challenge 4
*mcp_tools # live AWS docs tools
],
callback_handler=stream_callback,
system_prompt="..."
)

You: What is the interview process at Amazon and what should I prepare?

🔧 Using tool: recall_memory
🔧 Using tool: mcp_aws_documentation_search
🔧 Using tool: skill_gap_analyzer_tool

Agent: Amazon has 4 interview rounds:
1. Online Assessment (DSA — 2 problems in 90 mins)
2. Technical Round 1 (Problem Solving)
3. Technical Round 2 (System Design)
4. Bar Raiser (Behavioural + Leadership Principles)

Based on your profile (Python + SQL), focus on DSA first.
You are missing System Design and Java/C++ — here is your roadmap...