Languages
- Java
- Python
- SQL
- Bash
- JavaScript
- Node.js
- GraphQL
- Ruby
- C++
Dublin, Ireland / Open to opportunities
Prathamesh Kalamkar
Engineering
with intention.
Back-end engineer. Agent systems builder.
I turn complex ideas into reliable software.
Meet the work — and the twin that knows it.
+ THE ENGINEERING OBSERVATORY 01 / 10
pk.THE DIGITAL TWIN
06°15′ W
Which one is you?
MEET MY AGENTIC DIGITAL TWINCURIOUS? START HERE ↘
Ask his agentic digital twin what he’s built.
Grounded in my CV & public code. · Offers and commitments? Talk to me directly.
3.5+Years building back ends
10Public repositories
MScCybersecurity · First class
Dublin↗Based in Ireland
A little more about the human
01Open sourceBuilt in the openLess hand-waving.
Less hand-waving.
More source code.
Also completed
AWSNode.jsData Structures & AlgorithmsSystem Design
02ExperienceService history
Three and a half years of shipping back-end systems.
-
Jan 2025 → Present
Software Engineer, AI & Automation Systems
FreelanceLLM-powered back-end services for agentic workflows: tool orchestration, structured outputs, retries, validation and safety checks. Event-driven workflow engines that assemble context, invoke external tools and return reviewable output.
-
Oct → Dec 2024
Software Engineer Intern
Physics WallahBackend features in Java and Spring Boot across high-usage product workflows, plus debugging and release readiness.
-
Jun 2022 → Jul 2024
Software Engineer
matriXploitJava, Spring Boot, Python, SQL and REST APIs across shared production systems. Refactored roughly half of a legacy authentication module, and added diagnostics and health checks that cut regressions by about 35%.
-
Jan → Jun 2022
Technical Support Engineer
matriXploitRoot-cause analysis through SQL and logs, narrowing production issues to specific flows and escalating with clear findings.
Education
MSc Cybersecurity
Dublin Business School · first-class honours · 2024 to 2025
BSc Computer Science
Savitribai Phule Pune University · first class · 2021 to 2024
03CapabilitiesCapability registry
What he reaches for, and how deep it goes.
Back-end
- Spring Boot
- REST APIs
- Microservices
- Data modelling
- Auth flows
- Service contracts
- Structured logging
- JVM internals
AI & agents
- LLM integrations
- OpenAI APIs
- Claude
- LangChain
- LangGraph
- Agentic workflows
- Function calling
- Tool orchestration
- Retries & guardrails
- Prompt contracts
- MCP
Events & data
- Event-driven pipelines
- Context assembly
- Kafka
- SQS
- RAG
- Embeddings
- Vector stores
- Caching & rate limiting
Observability
- Logs, metrics, traces
- Run metadata
- Dashboards
- Debugging
- OpenTelemetry
- Datadog
Security & ops
- Input validation
- Prompt-injection defence
- Secret management
- Least-privilege design
- Integration testing
- CI/CD
- Docker
Production experience
Java
Spring Boot
Python
FastAPI
SQL
Docker
Git
GitHub
JavaScript
Bash
Claude
LangChain
LangGraph
Familiarity, stated as such
Node.js
GraphQL
Kafka
AWS
OpenTelemetry
Datadog
GitHub Actions
C++
Ruby
Solid = used in production work. Faded = familiarity, stated as such rather than overclaimed.
04Selected projectsDeployed workloads
Things he built to answer a question.
2026
Agentic Job Application Platform
A full-stack platform automating ATS workflows through agent-driven task execution, persistent state, human review gates and structured outputs. Tool execution layers with validation, retries and workflow controls keep long task chains inspectable.
PythonFastAPISQLAgent workflowsSupabase
2025
AI Agent Evaluation Harness
Compares agent responses, surfaces failure modes and improves orchestration through reproducible runs rather than intuition. Scoring, parsing and run metadata support iterative tuning of prompts, tool selection and error handling.
PythonTool callingStructured outputsPrompt evals
2025
SAFECode
An AI-assisted code review tool that analyses source for vulnerabilities and returns structured findings for triage, with clear input validation and predictable outputs so it fits into an existing developer workflow.
PythonFastAPISecurity analysisAI workflows
05AI systemsRetrieval pipeline
The retrieval is running on this page.
Answers here are not generated from a prompt and hoped for. A hybrid retriever scores every chunk of the CV corpus two ways: Okapi BM25 for term weighting, character-trigram cosine for the morphological variants BM25 misses. Reciprocal rank fusion then merges the two rankings by position, so an unbounded BM25 score cannot swamp a bounded cosine. Type anything and watch all three stages.
- Understands the hiring goal
- Builds the shortest safe plan
- Chooses only relevant public tools
- Verifies every claim before answering
Each answer includes the run: plan, evidence, tools used and verification outcome.
Model…
Groundingper claim
Repos10
SourcesCV + GitHub
goal → plan → tools → evidence → verify
Model Context Protocol
MCP servers
Agent systems
Built, not described.
06QuestionsControl plane
What people ask before they get in touch.
Is this actually accurate?
Every claim is retrieved from his CV or public repositories and cited underneath the answer. If something isn't evidenced, the agentic digital twin says so rather than guessing.
Can it negotiate or accept an offer?
No. It speaks about the work; it cannot commit anything on his behalf. Salary, offers, start dates and anything contractual are refused by design and handed to Prathamesh directly.
Why does it ask my name?
So it can point you at the most relevant work. It's optional. Skipping gives you the identical conversation.
What is it built on?
Python, FastAPI and SQLite with a provider-agnostic model layer. The source is public on GitHub, alongside the ten systems it talks about.
Public endpoint
Great systems start
Great systems start
with a conversation.
Have a role, a challenging problem, or a question about my work? Meet my agentic digital twin, or get in touch with the human behind it.