Aug 2023Education
Started at BITS Pilani
B.E. Computer Science Engineering — Pilani Campus
Open to SDE / AI engineering internships · willing to relocate
Tanush
Singhal
Build products, then prove they worked with real numbers.
0st
of 50+ teams
0+
REST APIs
0%
token cost cut
0
internships
Scroll
At a glance
The proof, up front.
Currently
Software Engineering Intern at Software Tree, building a 24/7 inventory-monitoring agent — remote from India to the USA.
Open to SDE / AI engineering internships · willing to relocate
0stof 50+ teamsFinTech hackathon
0+REST APIsshipped end-to-end
0%token cost cutgraph workloads
0internshipsUSA + Indian Railways
BITS Pilani — Pilani Campus
B.E. Computer Science Engineering
2023 — 2027
IIT Madras
B.S. Data Science & Applications (Online)
2023 — 2027
The short version
Ambiguous problems become shipped systems — across logistics, fintech, and AI.
01
Closing the loop between building and measuring.
A builder who finishes. In every project, the interesting part is the decisions — what to build, what to cut, and how to measure whether it moved the needle. Engineering is the unfair advantage; instrumentation is what stands out.
Across two internships and a hackathon win: event-driven services, 40+ REST APIs, a serverless voice assistant, and a published benchmark study. Impact isn't claimed — it's instrumented, defended with numbers, and written about in public.
Currently a remote Software Engineering Intern at Software Tree (USA), building a 24/7 inventory-monitoring agent on a JSON-RPC backend, while studying CS at BITS Pilani and Data Science at IIT Madras simultaneously.
BITS Pilani — Pilani Campus
B.E. Computer Science Engineering
2023 — 2027
IIT Madras
B.S. Data Science & Applications (Online)
2023 — 2027
Things I'm known for
- 01Published a 33-minute benchmark essay on the hidden cost of AI database access.
- 02Audited the railway system like a stranger would — the proudest feature turned out to be the biggest bottleneck.
- 03Removed the LLM from the monitoring loop after measuring that arithmetic beat aesthetics.
02
How I think
Engineering is a series of decisions, not a pile of features. Build the smallest honest version, measure whether it worked, and let the number — not the aesthetics — make the call.
“Rules where you need vigilance, AI where you need judgment.”
“Ask what breaks at 100× before shipping — and audit your own work like a stranger would.”
“The right unit is cost per convinced customer, not cost per minute.”
“When a false positive costs a career, AI advises and humans decide.”
03
Where I've shipped
Two internships building real, tested software — in the USA and for Indian Railways.
Software Engineering Intern
Oct 2025 — PresentSoftware Tree, LLC · USA · Remote
- Owned end-to-end design of a 24/7 inventory-monitoring service — an event-driven agent over a JSON-RPC backend that auto-detects low-stock items, computes reorder quantity, and dispatches alerts, replacing manual stock checks.
- Designed and ran a 20-session controlled benchmark across 4 tasks on GPT-5.4 + Claude-4.6, instrumenting tokens, cost and latency, and cut token (compute) cost by up to 58% for graph-shaped workloads.
- Built a reusable Python JSON-RPC client with Server-Sent Events streaming and session handling, giving the team typed, schema-safe data access in place of raw SQL.
- Raised release quality across 9 release iterations on live PostgreSQL / Oracle / MySQL / H2; root-caused a reproducible defect with a workaround and surfaced 2 JDBC blockers.
PythonLangChainLangGraphPostgreSQLJSON-RPCDocker
“Your work during the internship has been exceptional — not just technically strong, but also thoughtful, reliable, and mature in execution. The contributions you made around ORMCP, Gilhari, JDX, LLM systems, and agent workflows stood out, and several times you operated well beyond what I'd normally expect from an intern. Your work on the Inventory Watchdog project, the related YouTube video and podcast, and the benchmark article have been outstanding. More importantly, you approached problems with curiosity, ownership, and a strong learning mindset, which has made working with you a pleasure. I genuinely believe you have a very strong profile for roles in GenAI, backend infrastructure, AI systems engineering, and intelligent usage of AI tools, and I'm confident you'll do very well wherever you go next. You've made a real impact here.”
Software Development Intern
May 2025 — Jul 2025Centre for Railway Information Systems (CRIS) · India
- Designed and shipped a full-stack Parcel Management System end-to-end: 40+ REST APIs across 6 modules over a 5-table relational schema, JWT role-based access for 3 roles, and Docker Compose deployment with durable volumes.
- Engineered password-less OTP login with a 3-tier email failover (Gmail → Postmark → RapidAPI) that keeps authentication working when any single provider fails, plus a network-wide audit log.
Node.jsExpressSequelizeSQLiteJWTDocker ComposeReact
04
My timeline
The arc so far — from BITS & IIT Madras to shipping production systems and publishing in public.
Aug 2023Education
Enrolled at IIT Madras
B.S. Data Science & Applications (online), pursued alongside BITS
May 2025Work
CRIS internship begins
Full-stack Parcel Management System for Indian Railways
Jun 2025Milestone
1st of 50+ teams — LenDenClub hackathon
Led a 4-member team; shipped a serverless voice AI assistant in 48 hours
Jul 2025Project
Railway PMS shipped
40+ REST APIs, JWT auth, Docker Compose — capacity-planned for 7,000+ stations
Oct 2025Work
Software Tree internship
Remote SDE intern (USA) — building the Inventory Watchdog agent
Nov 2025Writing
Published Inventory Watchdog essay
Medium architecture deep-dive — why the LLM left the monitoring loop
Apr 2026Project
APOGEE 2026 — fraud detection system
Hybrid explainable scoring; AI advises, humans decide
Apr 2026Writing
Published MCP benchmark study
33-minute essay on the hidden tax of AI database access
PresentMilestone
Building in public
Shipping agentic backends, benchmarking LLM tooling, open to SDE / AI internships
05
Selected work
Each one ends with a decision principle — because the decision is what matters most.
06
The toolkit
Languages
- Python
- Java
- C/C++
- SQL
- JavaScript
- TypeScript
- Bash
Backend & Systems
- System Design
- Distributed Systems
- Microservices
- REST / WebSocket
- FastAPI
- Node / Express
- Concurrency
- DSA
Data & AI
- LLMs
- LangChain / LangGraph
- ReAct Agents
- RAG
- MCP
- Prompt Engineering
- LLM Evaluation
- scikit-learn
- pandas
Cloud & Tooling
- AWS (Lambda, API Gateway, Bedrock, S3)
- PostgreSQL
- Docker
- Git/CI-CD
- Linux
PythonJavaC/C++SQLJavaScriptTypeScriptBashSystem DesignDistributed SystemsMicroservicesREST / WebSocketFastAPINode / ExpressConcurrencyDSALLMsLangChain / LangGraphReAct AgentsRAGMCPPrompt EngineeringLLM Evaluationscikit-learnpandasAWS (Lambda, API Gateway, Bedrock, S3)PostgreSQLDockerGit/CI-CDLinuxPythonJavaC/C++SQLJavaScriptTypeScriptBashSystem DesignDistributed SystemsMicroservicesREST / WebSocketFastAPINode / ExpressConcurrencyDSALLMsLangChain / LangGraphReAct AgentsRAGMCPPrompt EngineeringLLM Evaluationscikit-learnpandasAWS (Lambda, API Gateway, Bedrock, S3)PostgreSQLDockerGit/CI-CDLinux
07
Coding activity
A year of building on GitHub, grinding LeetCode, and competing on Codeforces — one grid, color-coded by platform.
08
I write about the trade-offs
Long-form engineering essays, with companion video and podcast walkthroughs.
Achievements
- · 1st Place / 50+ Teams — Matrix Protocol AI Hackathon (LenDenClub, FinTech).
- · Technical Author on Medium — 2 long-form engineering essays with YouTube + podcast walkthroughs.
- · Shipped production code across two internships (USA + Indian Railways).
09
What people say
Reviews from the people I've shipped alongside — and the credentials to back the work.
Your work during the internship has been exceptional — not just technically strong, but also thoughtful, reliable, and mature in execution. The contributions you made around ORMCP, Gilhari, JDX, LLM systems, and agent workflows stood out, and several times you operated well beyond what I'd normally expect from an intern. Your work on the Inventory Watchdog project, the related YouTube video and podcast, and the benchmark article have been outstanding. More importantly, you approached problems with curiosity, ownership, and a strong learning mindset, which has made working with you a pleasure. I genuinely believe you have a very strong profile for roles in GenAI, backend infrastructure, AI systems engineering, and intelligent usage of AI tools, and I'm confident you'll do very well wherever you go next. You've made a real impact here.
10
Let's build something
Hiring, collaborating, or just curious?
Let's talk.
I'm open to sde / ai engineering internships · willing to relocate. Book a slot and my calendar handles the rest — or ask my AI twin anything first.