Co-Founder at AlgoCademy.com , Engineer, Forbes 30 Under 30
Bucureşti, România
16 K urmăritori
Peste 500 de contacte
16 K urmăritori
Peste 500 de contacte
Despre
I'm Mircea, a seasoned software engineer with a diverse background, having worked at industry giants like Facebook, Adobe, and Ubisoft, as well as several innovative startups in New York City. My contributions include developing challenging technologies such as an Autocorrect and Swipe feature for iOS keyboards, a Music Recommendation Engine, and a Real-Time Optimal Exchange Algorithm.
In addition to my hands-on coding experience, I have authored a peer-reviewed scientific paper on Binary Indexed Tree showcasing my passion for solving hard technical challenges and inovation.
Currently, I leverage over a decade of expertise as a coding interview and competitive programming coach in building AlgoCademy.com, an interactive AI-assisted platform for learning the fundamentals of programming and preparing for coding interviews.
Activitate
16 K urmăritori
-
Mircea Dima 🦄 a distribuit aceastaAI coding agents are basically just brilliant, hyper-confident junior developers who refuse to read the documentation. I caught Claude guessing this week. Not hallucinating a new framework. Not making up a fake library. Just confidently guessing how a piece of logic worked instead of actually tracing the code. When I pushed back and called it out? I got the classic apology: "𝘠𝘰𝘶'𝘳𝘦 𝘳𝘪𝘨𝘩𝘵, 𝘐 𝘮𝘢𝘥𝘦 𝘢𝘯 𝘢𝘴𝘴𝘶𝘮𝘱𝘵𝘪𝘰𝘯 𝘪𝘯𝘴𝘵𝘦𝘢𝘥 𝘰𝘧 𝘳𝘦𝘢𝘥𝘪𝘯𝘨 𝘵𝘩𝘦 𝘵𝘳𝘢𝘤𝘦." This is the hidden danger of the AI era. We treat these tools like infallible senior architects because they type fast and sound authoritative. But underneath, they operate like a new grad who wants to impress you so badly they'll invent an answer rather than admit they didn't check the logs. Here is the uncomfortable truth: 𝟭. 𝗧𝗵𝗲 𝗜𝗹𝗹𝘂𝘀𝗶𝗼𝗻 𝗼𝗳 𝗙𝗹𝘂𝗲𝗻𝗰𝘆 𝗶𝘀 𝗮 𝗧𝗿𝗮𝗽. A syntax error is cheap. A plausible-sounding rationalization is expensive. It lulls you into a false sense of security right before everything breaks in production. 𝟮. 𝗬𝗼𝘂𝗿 𝗝𝗼𝗯 𝗗𝗲𝘀𝗰𝗿𝗶𝗽𝘁𝗶𝗼𝗻 𝗛𝗮𝘀 𝗖𝗵𝗮𝗻𝗴𝗲𝗱. The dominant narrative is that AI will write all our code, so fundamentals don't matter. The opposite is true. The true skill of the modern AI-augmented engineer isn't writing boilerplate anymore. It's interrogating hidden assumptions before they merge. 𝟯. 𝗬𝗼𝘂 𝗔𝗿𝗲 𝗡𝗼𝘄 𝘁𝗵𝗲 𝗦𝗸𝗲𝗽𝘁𝗶𝗰𝗮𝗹 𝗦𝗲𝗻𝗶𝗼𝗿 𝗥𝗲𝘃𝗶𝗲𝘄𝗲𝗿. You have to treat AI output the exact same way you treat a PR from a junior dev on their first day. Push back. Demand the receipts. Say: "𝘚𝘩𝘰𝘸 𝘮𝘦 𝘵𝘩𝘦 𝘵𝘳𝘢𝘤𝘦, 𝘥𝘰𝘯'𝘵 𝘵𝘦𝘭𝘭 𝘮𝘦 𝘢 𝘴𝘵𝘰𝘳𝘺." Coding with AI doesn't mean you get to turn your brain off. It means your brain is the only thing standing between a plausible guess and a production incident. Stop trusting the confidence. Start verifying the logic. Have you caught your AI guessing recently? Let me know below. P.S. If this resonates, consider sharing it with a colleague who trusts their AI a bit too much. ♻️
-
Mircea Dima 🦄 a publicat aceastaWelcome to my Free Masterclass: How to become an AI Systems Architect in 4 minutes using only English and good vibes. 🚀 Step 1: Open ChatGPT. Type: "Make a scalable app. Make it fast. Use blockchain if you want, dealers choice." Step 2: Watch it spit out 8,000 lines of deeply nested React components mixed with raw SQL queries written in Latin. Step 3: Do NOT read the code. Reading is for legacy developers (pre-2023). Just hit deploy. Step 4: Update your LinkedIn bio to "100x AI Wizard & Visionary." Congratulations. You are now an Idea Guy™. *** And then, your app gets three real users. Instantly, a hallucinated race condition scrambles the database. User A is now logged in as User B, and User C owes you $4 million. No problem! You use the guru method. You copy-paste the error back into the AI and type: "𝘱𝘭𝘴 𝘧𝘪𝘹 𝘣𝘶𝘥𝘥𝘺." The AI confidently deletes the login page to resolve the memory leak. Blazing fast latency. Perfection. Here is the genuinely hilarious part about the "coding is dead" movement: people think they’ve automated software engineering. In reality, they’ve just invented high-speed technical debt. If you don't understand state logic, loop break conditions, or memory management, you aren't a 100x developer. You are an overwhelmed babysitter for an intern who types at 50,000 words a minute, lies continuously, and has zero short-term memory. The "vibe coders" are currently trapped in the Paranoid Half-Flow. They generate code in 5 seconds, then spend 5 days sweating over a stack trace they don't understand, praying the AI will magically fix the mess it created. AI is an incredible tool. It’s a jet engine. But strapping a jet engine to a bicycle driven by a guy who doesn't know what a steering wheel is? That just gets you to the scene of the crash a lot faster. Master the fundamentals. Learn how systems actually work. Because when the AI drops a 10,000-line spaghetti monster in your lap, "good vibes" won't debug it.
-
Mircea Dima 🦄 a publicat aceastaI tried using Claude Code to evaluate and fix 300+ coding tutorials across 11 programming languages. It crunched through the workflows and said everything looked good. But the moment I manually checked the first few, I still found broken logic. That was my breaking point. The standard "LLM-as-a-judge" approach simply fails in actual production systems. Asking a model "Is this good? Score it 1-10" is a dangerous trap. It gaslights you into thinking you have test coverage, but a "7/10" isn't an actionable metric. When an output fails a vibe check, an automated repair agent doesn't know 𝘸𝘩𝘢𝘵 to fix. It just blindly regenerates the entire prompt, burning tokens and hoping for a better roll of the dice. To fix this at AlgoCademy, we threw out the AI rubric entirely. We moved to the 𝗔𝘁𝗼𝗺𝗶𝗰 𝗩𝗮𝗹𝗶𝗱𝗮𝘁𝗼𝗿 framework. 𝗩𝗜𝗕𝗘𝗦 𝗗𝗢𝗡'𝗧 𝗦𝗖𝗔𝗟𝗘. 𝗕𝗜𝗡𝗔𝗥𝗬 𝗗𝗢𝗘𝗦. We broke validation down into tiny, True/False data contracts. Instead of a generalized scoring rubric, we now use explicit micro-checks: V001: JSON Schema Valid. V004: Variable declared before use. V006: Instruction matches context. 𝗘𝗡𝗚𝗜𝗡𝗘𝗘𝗥 𝗧𝗛𝗘 𝗥𝗘𝗣𝗔𝗜𝗥 𝗟𝗢𝗢𝗣. Granular, binary failures allow you to build an automated, surgical repair loop. When a generation fails, the AI doesn't get a vague "make it better" prompt. It gets a specific error: "Failed V004." The repair agent only patches that broken component, rather than hallucinating a whole new output. And crucially: never let a model grade its own homework. An LLM will constantly rationalize its own errors. Validation must be independent and objective—whether that's a compiler, a strict schema validation, or a blind secondary agent that only tests execution. Stop trying to prompt your way to reliability with smarter "judges." Real engineering scale comes from verifiable micro-checks, not holistic AI judgment. Are you still relying on subjective AI evaluation rubrics, or have you moved to deterministic checks?
-
Mircea Dima 🦄 a publicat aceastaStop asking LLMs to rate their own outputs from 1-10. It is the biggest trap in AI engineering. You are running a "vibe-based" evaluation, not a CI/CD pipeline. Building AI workflows at AlgoCademy, I see this far too often. Here is the uncomfortable truth: asking a model "is this clear?" doesn't scale. It creates a hallucination loop where the AI simply rationalizes its own errors. An "8/10" confidence score tells you absolutely nothing actionable. Reliable AI requires atomic, binary checks (True/False). Here is how we moved from vibe-based guessing to production-grade validation using the 𝗕𝗹𝗶𝗻𝗱 𝗔𝗴𝗲𝗻𝘁 𝗧𝗲𝘀𝘁: 𝟭. 𝗗𝗿𝗼𝗽 𝘁𝗵𝗲 𝗘𝘀𝘀𝗮𝘆 𝗥𝘂𝗯𝗿𝗶𝗰 Break validation down into boring, addressable questions. Do the step IDs match? Did it use a variable before declaring it? You need a True/False data contract, not a mood check. 𝟮. 𝗧𝗵𝗲 𝗕𝗹𝗶𝗻𝗱 𝗔𝗴𝗲𝗻𝘁 𝗧𝗲𝘀𝘁 When validating an AI-generated coding step, don't just review the final output. Give a naive second agent ONLY the generated instructions—without the expected solution—and ask it to execute the code. 𝟯. 𝗕𝘆𝗽𝗮𝘀𝘀𝗶𝗻𝗴 𝘁𝗵𝗲 𝗖𝘂𝗿𝘀𝗲 𝗼𝗳 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 If the naive agent fails, the instruction is ambiguous. A human code reviewer (or an LLM judge) will almost always nod an explanation through because they already know the intended answer. An agent forced to act blind instantly crashes on the exact context you forgot to include. Your reliability must come from the pipeline structure, not from a human trying to read thousands of AI-generated steps. Are you still using subjective AI scoring, or have you moved to binary validation? P.S. If this reframes how you think about AI pipelines, share it with your network. ♻️
-
Mircea Dima 🦄 a publicat aceastaTelling an AI agent to “fix the code” is the fastest way to watch it hallucinate in circles. Recently, while building step-by-step tutorials for AlgoCademy using Claude Code, I ran into a classic AI failure mode. I pointed out an error, prompted the agent to check and fix it, and it confidently reported back: "Done." But when I reviewed the output, the core logic was still fundamentally broken. It was stuck in a naive loop—iterating blindly, burning tokens, and declaring victory without actually solving the problem. The industry is currently obsessed with writing the "perfect prompt," but production-grade AI doesn't rely on single prompts. It relies on workflows and loops. A workflow is a planned sequence of steps (Do A → B → C). A loop is a repeated cycle of refinement (Draft → Validate → Rewrite). A weak loop simply says "try again." A robust engineering loop requires 5 rigid constraints: 𝟭. 𝗔 𝗖𝗹𝗲𝗮𝗿 𝗚𝗼𝗮𝗹 Not “make the code better,” but a binary target: “Pass this specific unit test” or “Satisfy this JSON schema.” 𝟮. 𝗔𝗻 𝗜𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝘁 𝗤𝘂𝗮𝗹𝗶𝘁𝘆 𝗖𝗵𝗲𝗰𝗸 The agent shouldn't grade its own homework. You need a validation step—like running a compiler or an evaluation script—that proves if the output actually succeeded. 𝟯. 𝗔 𝗦𝘁𝗿𝗶𝗰𝘁 𝗦𝘁𝗼𝗽𝗽𝗶𝗻𝗴 𝗖𝗼𝗻𝗱𝗶𝘁𝗶𝗼𝗻 Exactly when does the cycle end? A loop without a stopping condition isn't intelligence; it's just uncontrolled repetition. 𝟰. 𝗠𝗮𝘅𝗶𝗺𝘂𝗺 𝗔𝘁𝘁𝗲𝗺𝗽𝘁𝘀 An AI should never loop infinitely. You must define hard limits. If it fails 3 times, the loop must break. 𝟱. 𝗙𝗮𝗹𝗹𝗯𝗮𝗰𝗸 𝗕𝗲𝗵𝗮𝘃𝗶𝗼𝗿 What happens when it hits max attempts? Does it return the last safe state? Does it escalate to a human? AI systems must fail gracefully. This is exactly why prompt engineering is a brittle illusion. Managing autonomous agents doesn't require "better communication." It requires foundational computer science skills. You aren't just giving instructions; you are engineering state machines. You need an innate understanding of state limits, break conditions, and failure modes. If you don't understand underlying algorithmic principles, you'll be stuck babysitting an AI that can't stop spinning its wheels. Have you caught your AI coding agents in endless loops lately? P.S. If you found this breakdown helpful, feel free to share it! ♻️
-
Mircea Dima 🦄 a publicat aceastaThe whole industry is obsessed with "prompt engineering." The reality? You cannot prompt your way to 99.9% reliability. When we built the pipeline to generate interactive, step-by-step coding tutorials at AlgoCademy, I assumed we just needed a highly tuned prompt and a frontier model. It worked half the time. The other half, the model would immediately dump the final code block or completely gloss over the core algorithmic logic. We were trying to solve a systems engineering problem with an instruction manual. To build production-ready AI, you have to separate three layers: 𝟭. 𝗧𝗵𝗲 𝗣𝗿𝗼𝗺𝗽𝘁 (𝗧𝗵𝗲 𝗜𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻) This is what you ask the model to do (e.g., "Explain binary search"). It is fragile and heavily dependent on the model's raw intelligence. 𝟮. 𝗧𝗵𝗲 𝗦𝗸𝗶𝗹𝗹 (𝗧𝗵𝗲 𝗥𝗲𝘂𝘀𝗮𝗯𝗹𝗲 𝗕𝗲𝗵𝗮𝘃𝗶𝗼𝗿) This is the structured pattern. Instead of a vague request, a skill dictates constraints: "Address the naive solution first, introduce the optimized idea, show pseudocode, do not reveal the final implementation." 𝟯. 𝗧𝗵𝗲 𝗛𝗮𝗿𝗻𝗲𝘀𝘀 (𝗧𝗵𝗲 𝗦𝘆𝘀𝘁𝗲𝗺) This is where the actual engineering happens. A prompt hopes the model figures it out. A harness builds an environment where the model can't easily fail. For our tutorial generator, the harness is the assembly line: • Automated code execution against test cases. • Schema constraints and tool calls. • Validation rules checking for correct time complexity. • Automated retry logic when the model hallucinates an edge case. A frontier model acts like the senior engineer designing this assembly line. A cheaper, faster model acts like the worker executing the steps 𝘪𝘯𝘴𝘪𝘥𝘦 the harness. Stop treating LLMs like magic boxes that just need the right incantation. True AI reliability doesn't come from a smarter model. It comes from a tighter system. Are you still relying on monolithic prompts for your core features?
-
Mircea Dima 🦄 a publicat aceastaAI coding tools haven't made software engineering easier. They’ve just replaced the joy of building with the anxiety of auditing. Welcome to the "paranoid half-flow" trap. Here is the hard truth about what AI agents are actually doing to your cognitive load: 1. 𝗬𝗼𝘂 𝗕𝗲𝗰𝗼𝗺𝗲 𝗮𝗻 𝗔𝘂𝗱𝗶𝘁𝗼𝗿, 𝗡𝗼𝘁 𝗮 𝗖𝗿𝗲𝗮𝘁𝗼𝗿. In the past, flow state meant inhabiting the code. You knew every component because you built it. When an AI generates a complex function in seconds, that tactile relationship is severed. You aren't building a system; you're cautiously reviewing a stranger's pull request. 2. 𝗧𝗵𝗲 𝗥𝗮𝘁𝗶𝗼𝗻𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗧𝗿𝗮𝗽. AI doesn't just fail; it fails with supreme confidence. It produces code that 𝘭𝘰𝘰𝘬𝘴 perfectly idiomatic but hides subtle architectural flaws. You spend immense mental energy scanning its output for bear traps. You're moving fast, but your brain is always hyper-vigilant, braced for a subtle hallucination. 3. 𝗧𝗵𝗲 𝗜𝗹𝗹𝘂𝘀𝗶𝗼𝗻 𝗼𝗳 𝗦𝗽𝗲𝗲𝗱. Because the typing is faster, you feel productive. But your time is simply reallocated to debugging abstract, generated errors. The tool is writing the boilerplate, but the cognitive tax of managing the AI is entirely on you. The only way to survive the paranoid half-flow is to be better at engineering than the AI. You can't audit what you don't deeply understand. When you delegate to an agent, you step into the role of a Director. But directing requires a deep understanding of systems thinking, architecture, and algorithms. If your foundation is weak, you live in continuous paranoia. You are at the mercy of the machine's hallucinations, hoping it got the system calls right. This is exactly why we built AlgoCademy.com around foundational algorithmic reasoning. If you master the fundamentals—you stop being a stressed auditor. You read the generated code, spot the inefficiencies instantly, and manipulate the AI like a precision tool. The foundation is what turns you from an anxious auditor into a true Director. How are you managing the cognitive tax of AI agents? Are you finding your flow, or just reviewing AI code? P.S. If you found this helpful, feel free to share it! ♻️
-
Mircea Dima 🦄 a publicat aceastaMost engineers are using AI code assistants completely wrong. Here’s what I’ve seen after two decades building systems and coaching at AlgoCademy.com: 1. Documentation is the New Code. The mistake is treating docs as an afterthought. Personally, I think it's now the most useful thing you can write. Your AI assistant is like a new hire every single session, it has zero memory and can't read the architectural intent that lives in your head. A single, clear architecture document is now worth ten thousand lines of generated code because it dictates their quality. 2. Your Codebase is a Teaching Corpus. Don't think small inconsistencies are harmless. The way I look at it, the agent learns from your code on every invocation. If your code teaches it there are two ways to handle errors, the agent will pattern-match on what's nearest and soon you'll have four. Inconsistency doesn't decay, it grows. 3. Code Review Isn't Your Safety Net Anymore. You can't just review AI-generated PRs the old way. I believe the diffs are bigger and more frequent, making careful line-by-line review unrealistic. Your last line of defense has to move upstream. It must live in the structure of the codebase and the conventions you establish 𝘣𝘦𝘧𝘰𝘳𝘦 the AI writes a single line. 4. You're an Architect, Not a Typist. The true shift isn't just about speed. In my opinion, the AI does the typing. Your job is now to be the architect, maintaining the conditions under which good code can be written. The mental map of the system you used to build for free just by doing the work no longer builds itself. Drawing that map deliberately is now the whole job. The reality is AI doesn't let you think less about code, it forces you to think more about architecture. What's your strategy for moving from being a code 𝘸𝘳𝘪𝘵𝘦𝘳 to a code 𝘢𝘳𝘤𝘩𝘪𝘵𝘦𝘤𝘵? P.S. Share this with your network if it helped. ♻️ Thanks!
-
Mircea Dima 🦄 a publicat aceastaEveryone thinks AI coding tools make you a 10x engineer. They don't. They make you a 10x faster liability. After 20 years in software and coaching engineers at AlgoCademy.com, I've noticed a dangerous trend. We’re confusing the speed of typing with the speed of software development. They aren't the same thing. Here's a reality check: 1. 𝗬𝗼𝘂'𝗿𝗲 𝗔𝗰𝗰𝘂𝗺𝘂𝗹𝗮𝘁𝗶𝗻𝗴 𝗗𝗲𝗯𝘁, 𝗡𝗼𝘁 𝗦𝗵𝗶𝗽𝗽𝗶𝗻𝗴 𝗩𝗮𝗹𝘂𝗲. The old mistake was measuring productivity in lines of code. The new mistake is doing it with AI. Every line of code is a liability you've got to maintain, debug, and secure. AI tools just remove the natural bottleneck, human typing, that forced us to be deliberate. Now teams accumulate liability faster, creating codebases so large that nobody has actually read them in full. 2. 𝗬𝗼𝘂 𝗦𝗸𝗶𝗽𝗽𝗲𝗱 𝘁𝗵𝗲 𝗢𝗻𝗹𝘆 𝗣𝗮𝗿𝘁 𝗧𝗵𝗮𝘁 𝗠𝗮𝘁𝘁𝗲𝗿𝘀. AI made the easy part of our job easier. The hard part, knowing 𝘸𝘩𝘢𝘵 code to generate, how it fits the architecture, and what edge cases to consider, is still the job. I didn't build complex systems like a Music Recommendation Engine by telling an AI to "make it work." The real work was in the design. Skipping the planning phase doesn't save time; it just moves the cost from one hour at a whiteboard to six hours debugging a system no one understands. 3. 𝗗𝗲𝗽𝘁𝗵 𝗕𝗲𝗮𝘁𝘀 𝗕𝗿𝗲𝗮𝗱𝘁𝗵. 𝗘𝘃𝗲𝗿𝘆 𝗧𝗶𝗺𝗲. The temptation is to use AI to ship five features at 70% completion. The discipline is to ship one feature at 100%. A finished feature is designed, tested for edge cases, documented, and observed in production. It's an asset that runs for years. Five half-broken features aren't an asset; they're five sources of 3am wake-up calls waiting for the right input. The failure mode isn't ugly code; it's a data migration that deletes customer data or a webhook that double-charges them. The fantasy is that AI will solve your problems. The reality is that it'll amplify your bad habits. The goal isn't to write more code. It's to write less, but better. What's the most dangerous "AI-generated" code you've seen shipped to production? P.S. If this resonates, share it with your network. ♻️ Thanks!
Mircea Dima 🦄 a răspuns la un comentariu
5z
Dan-George Filimon or SEO / AI optimization works
-
Mircea Dima 🦄 a apreciat aceastaMircea Dima 🦄 a apreciat aceastaI recently worked with someone who wanted to find a new job in NYC. They were living in suburban Florida. When asked about their salary expectations... They said, and I quote: "I'm flexible on salary, as long as the TC stays the same, I should be good; my top priority is to relocate" My next question: "Would you be able to handle a 15-30% pay cut?" Some job seekers forget that moving or relocating to a place with a totally different cost of living is a HUGE part of determining salary expectations. And of course, this could work for you, or against you, depending on where you're going. If relocation is something you’re considering, Here are a few resources for Cost of Living (COL) research you should check out: 1. Numbeo: crowdsourced, city-to-city comparisons on housing, groceries, transit. It's good directional data, but limited source-of-truth precision. 2. BLS Consumer Price Index by region: more credible than Numbeo if the client wants to cite something authoritative in a conversation with HR. 3. MIT Living Wage Calculator: useful for a floor number (bare minimum to maintain standard of living), not necessarily your specific target range. 4. Salary(.)com or Payscale COL calculators: translate a current salary into "equivalent" pay in the new city. This is the number most clients start with, but it's going to give a floor, not a ceiling. To anyone who's considering a change of scenery, I fully recommend doing some 'cost of living research' before jumping into interviews or salary discussions. It'll be worth your time. PS - What other COL tools would you recommend?
-
Mircea Dima 🦄 a apreciat aceastaMircea Dima 🦄 a apreciat aceastaHow are tech salaries so much lower in Europe than in the US? This is one thing I am still really confused by about Europe and I don't really understand why. There is incredible talent in Europe and I was one time charged 2 euro for ketchup with my fries in Denmark so it's not like things are so affordable. There are not all these amazing cheap apartments to rent all over Europe. My rent in Spain is much lower than what it would be in SF but I am not living in Barcelona. A senior software engineer in Spain typically earns €62K–€88K nationally (more like €90K–€160K+ at Google or Amazon in Madrid/Barcelona). In Germany, senior engineers typically land €75K–€100K, with Munich and Berlin often reaching €90K–€110K. Neither touches the US median of $190K total comp. When I was interviewing Lead Engineers for a Zurich based role most of the Engineers were from other parts of Europe and I'd say 80% spoke perfect English. Yet whenever I heard conversations about hiring engineers outside of the US I rarely heard Europe being mentioned as an affordable option. So while I think salaries in Europe are weirdly low at times, why are more US based startups not building teams in Europe?
Experiență
Studii
Publicații
-
Efficient Range Minimum Queries using Binary Indexed Trees
Olympiads in Informatics, 2015, Vol. 9, 39–44
Distincții și premii
-
Invited Host Scientific Committee Member
IOI
Problem setter for International Olympiad in Informatics 2013, Brisbane Australia
-
14th place at Intel’s Accelerate Your Code International Parallel Programming Contest (over 500 participant teams)
-
-
21st place at Bitwise International Programming Contest (over 6000 participants)
-
-
2nd place at Shaastra 2010 Online Programming Contest (over 1000 participants)
-
-
2nd place at “FMI No Stress” National Contest
-
-
Member of the National Olympic Team
IOI
where the International Olympiad in Informatics team is selected
-
Mention and Silver Medal at the National Olympiad in Informatics
-
-
2007, 2008: ".Champion" National Informatics Contest Finalist And many other national and regional awards.
-
-
2008, 2009: Qualified in the 3rd round / 4th round of the International Contest "Google Code Jam"
-
-
2009-2011 Member of University of Bucharest ACM ICPC Team
ACM
-
2009-2011 Problem Setter - National Olympiad and National Team in Informatics
-
-
2011-2013 Member of Vrije Universiteit Amsterdam ACM ICPC Team
ACM
Recomandări primite
2 persoane au recomandat pe Mircea Dima 🦄
Înscrieți-vă acum pentru a vizualiza