Kashish Kumar

@asmah2107 Not everybody needs to build a harness

Backend Engineering is the hidden engine of scalable AI. If you are an ML Engineer, Go deeper: > gRPC & Protobufs (Low latency payloads) > AsyncIO & Event Loops (Non-blocking concurrency) > Connection Pooling (DB efficiency) > Kafka & CDC (Real-time data ingestion) > Backpressure Handling (Surviving inference bursts) > Redis Pipelining (Batching feature fetch) > Vector Indexing (HNSW/IVF in Qdrant) > Graph Traversal (Cypher in Memgraph) > Semantic Caching (Saving GPU compute) > Distributed Locking (Race condition safety) > Rate Limiting (Token buckets for LLMs) > Idempotency (Safe retries) > Database Sharding (Horizontally scaling state) > OpenTelemetry (Tracing request lifecycles) Architecture > Algorithms.

writing a CLI is underrated product work. a good CLI turns a 5-step dashboard workflow into a one-liner. developers will love your product more if they never have to open your dashboard.

@keshavchan To learn from it

i do not understand why people dwell on the past and like to query their past selves. you should be spending all of your time trying to go forwards and improve your current ways of seeing the world. we are living in the most interesting of times. we have, i think for the first time ever, access to magical powers. let's allow the past to be retro and dusty and give space for the present to shine?

@zarazhangrui Can you share ?

I hoard X bookmarks and never read them. So I built an extension that injects a bookmarked post into my main feed every time I open X (almost like an ad). Now I read my bookmarks. The trick was hijacking real estate I already check 50 times a day

@icanvardar In Enterprises, gemini is used all the time for better quality to cost ratio

google already lost the llm race no one i know uses gemini anymore, i don’t even hear it mentioned that often

@adxtyahq Sounds like a good plan

random thought i feel the chinese government helps AI labs with GPU’s to keep model costs extremely low so more users = more data more data = better models better models = more products built on top over time you end up with an entire ecosystem and a massive advantage, maybe the goal was never the model itself, but everything that gets built around it just a theory

@shreyasico Cool things happen where cool people live

why do all cool things happen in bangalore?

@jonathan_wilke Human computer interactions is evolving from clicks to natural language

I don't get the hype around CLI coding tools like Claude Code. Human-computer interaction evolved past the terminal 30 years ago for a reason. UIs won. Why are we regressing?

@akshushh @truefoundry thanks Akshat

@Kashish__Kumar_ @truefoundry added to jobs in the wild: jobsinthewild.tal.af/j/974 :)

We're hiring Forward Deployed Engineers (GTM) at @truefoundry The job: follow the full arc from discovery to architecture to POC to close across multiple enterprise accounts at once. Right now, that's one person. Our CTO walks in, runs discovery, designs the architecture, owns the POC, and closes the technical side. Works every time. The problem: it's not documented, not repeatable, and lives entirely in his head. We want the repeatable version of that. If you've done this, or you're a backend engineer who wants to — this is the role.

@aj_livess @truefoundry yes, check out careers page

@Kashish__Kumar_ @truefoundry You guys offer internships?? I’m still in college but looking for full time roles where I can apply myself and see some tangible results linkedin.com/in/aj-codess?u…

@rajatendu_dey @truefoundry great

@Kashish__Kumar_ @truefoundry Found the role interesting, applied 😁

@truefoundry app.careerpuck.com/job-board/true…

@O_Anu_O Gemini is better at images, and you can use OCR specific vision language models like Paddle OCR or Landing AI

Which model is the best in terms of cost and quality for layout-aware OCR extraction on pdfs? PDFs can be complex with embedded images ( images scanned as PDFs) I'm currently using GPT 4.1 but would love to hear if you've used any LLM model in ENTERPRISE and PRODUCTION!!

@merishabh_singh @truefoundry is hiring

COMPANIES THAT JUST RAISED $10-50M (NOW #HIRING) All 22 of these companies are hiring and backed by top funds...

Increasingly, I believe companies may need to be rebuilt from the ground up, where you have a single timeline of all observability + product metrics + file changes laid out in a retrievable system, like Datadog + Posthog + Google Drive + Slack (really unified filesystem of Claude Code chats + Codex chats). This might be the new data foundation for any and all companies to maximize AI. Needs to be rebuilt because keeping track of diffs on existing system basically impossible to produce longitudinal information on decisions and rollbacks, something coding agent storage companies are actively trying to figure out, but this should extend to businesses as a whole. Highly skeptical existing businesses will adopt this though because it means overhauling everything about their instrumentation and business data, but I think businesses built on this foundation probably can execute 100x better and faster

Pitch: every company should have one day a month where the only job is upskilling with AI. Not generic training. Teams exploring “how can AI help ME do what I already do better?” Imagine the compounding effect over a year. Why doesn’t this exist?

@driscollis uv

What's your favorite #Python packaging tool and why? 🐍 uv 🐍 Poetry 🐍 Pipenv 🐍 Hatch 🐍pipenv Or something else?

@adxtyahq Seems obvious

Before learning agents, MCPs, workflows, and frameworks, learn embeddings Embeddings power semantic search, RAG, recommendations, clustering, retrieval, memory systems, and a lot more. It's one of the highest-ROI concepts in AI Engineering rn Some resources: • "Neural Networks: Zero to Hero" by Andrej Karpathy • "Let's build GPT" by Andrej Karpathy • "Deep Dive into LLMs like ChatGPT" by Andrej Karpathy @karpathy is genuinely one of the best teachers for understanding what's happening under the hood instead of just calling APIs